Showing 1–50 of 315 results for author: Choi, H

Does Incomplete Syntax Influence Korean Language Model? Focusing on Word Order and Case Markers

Authors: Jong Myoung Kim, Young-Jun Lee, Yong-jin Han, Sangkeun Jung, Ho-Jin Choi

Abstract: Syntactic elements, such as word order and case markers, are fundamental in natural language processing. Recent studies show that syntactic information boosts language model performance and offers clues for people to understand their learning mechanisms. Unlike languages with a fixed word order such as English, Korean allows for varied word sequences, despite its canonical structure, due to case m… ▽ More Syntactic elements, such as word order and case markers, are fundamental in natural language processing. Recent studies show that syntactic information boosts language model performance and offers clues for people to understand their learning mechanisms. Unlike languages with a fixed word order such as English, Korean allows for varied word sequences, despite its canonical structure, due to case markers that indicate the functions of sentence components. This study explores whether Korean language models can accurately capture this flexibility. We note that incomplete word orders and omitted case markers frequently appear in ordinary Korean communication. To investigate this further, we introduce the Syntactically Incomplete Korean (SIKO) dataset. Through SIKO, we assessed Korean language models' flexibility with incomplete syntax and confirmed the dataset's training value. Results indicate these models reflect Korean's inherent flexibility, accurately handling incomplete inputs. Moreover, fine-tuning with SIKO enhances the ability to handle common incomplete Korean syntactic forms. The dataset's simple construction process, coupled with significant performance enhancements, solidifies its standing as an effective data augmentation technique. △ Less

Submitted 12 July, 2024; originally announced July 2024.

Comments: COLM 2024; Code and dataset is available in https://github.com/grayapple-git/SIKO

Information-theoretic classification of the cutoff phenomenon in Markov processes

Authors: Youjia Wang, Michael C. H. Choi

Abstract: We investigate the cutoff phenomenon for Markov processes under information divergences such as $f$-divergences and Rényi divergences. We classify most common divergences into four types, namely $L^2$-type, $\mathrm{TV}$-type, separation-type and $\mathrm{KL}$ divergence, in which we prove that the cutoff phenomenon are equivalent and relate the cutoff time and window among members within each typ… ▽ More We investigate the cutoff phenomenon for Markov processes under information divergences such as $f$-divergences and Rényi divergences. We classify most common divergences into four types, namely $L^2$-type, $\mathrm{TV}$-type, separation-type and $\mathrm{KL}$ divergence, in which we prove that the cutoff phenomenon are equivalent and relate the cutoff time and window among members within each type. To justify that this classification is natural, we provide examples in which the family of Markov processes exhibit cutoff in one type but not in another. We also establish new product conditions in these settings for the processes to exhibit cutoff, along with new results in non-reversible or non-normal situations. The proofs rely on a functional analytic approach towards cutoff. △ Less

Submitted 9 July, 2024; originally announced July 2024.

Comments: 55 pages

MSC Class: 60J05; 60J10; 60J25; 60J27; 94A15; 94A17

Empirical Study of Symmetrical Reasoning in Conversational Chatbots

Authors: Daniela N. Rim, Heeyoul Choi

Abstract: This work explores the capability of conversational chatbots powered by large language models (LLMs), to understand and characterize predicate symmetry, a cognitive linguistic function traditionally believed to be an inherent human trait. Leveraging in-context learning (ICL), a paradigm shift enabling chatbots to learn new tasks from prompts without re-training, we assess the symmetrical reasoning… ▽ More This work explores the capability of conversational chatbots powered by large language models (LLMs), to understand and characterize predicate symmetry, a cognitive linguistic function traditionally believed to be an inherent human trait. Leveraging in-context learning (ICL), a paradigm shift enabling chatbots to learn new tasks from prompts without re-training, we assess the symmetrical reasoning of five chatbots: ChatGPT 4, Huggingface chat AI, Microsoft's Copilot AI, LLaMA through Perplexity, and Gemini Advanced. Using the Symmetry Inference Sentence (SIS) dataset by Tanchip et al. (2020), we compare chatbot responses against human evaluations to gauge their understanding of predicate symmetry. Experiment results reveal varied performance among chatbots, with some approaching human-like reasoning capabilities. Gemini, for example, reaches a correlation of 0.85 with human scores, while providing a sounding justification for each symmetry evaluation. This study underscores the potential and limitations of LLMs in mirroring complex cognitive processes as symmetrical reasoning. △ Less

Submitted 8 July, 2024; originally announced July 2024.

Comments: Accepted in Future Technology Conference (FTC) 2024

How DNNs break the Curse of Dimensionality: Compositionality and Symmetry Learning

Authors: Arthur Jacot, Seok Hoan Choi, Yuxiao Wen

Abstract: We show that deep neural networks (DNNs) can efficiently learn any composition of functions with bounded $F_{1}$-norm, which allows DNNs to break the curse of dimensionality in ways that shallow networks cannot. More specifically, we derive a generalization bound that combines a covering number argument for compositionality, and the $F_{1}$-norm (or the related Barron norm) for large width adaptiv… ▽ More We show that deep neural networks (DNNs) can efficiently learn any composition of functions with bounded $F_{1}$-norm, which allows DNNs to break the curse of dimensionality in ways that shallow networks cannot. More specifically, we derive a generalization bound that combines a covering number argument for compositionality, and the $F_{1}$-norm (or the related Barron norm) for large width adaptivity. We show that the global minimizer of the regularized loss of DNNs can fit for example the composition of two functions $f^{*}=h\circ g$ from a small number of observations, assuming $g$ is smooth/regular and reduces the dimensionality (e.g. $g$ could be the modulo map of the symmetries of $f^{*}$), so that $h$ can be learned in spite of its low regularity. The measures of regularity we consider is the Sobolev norm with different levels of differentiability, which is well adapted to the $F_{1}$ norm. We compute scaling laws empirically and observe phase transitions depending on whether $g$ or $h$ is harder to learn, as predicted by our theory. △ Less

Submitted 8 July, 2024; originally announced July 2024.

Topological Persistence Guided Knowledge Distillation for Wearable Sensor Data

Authors: Eun Som Jeon, Hongjun Choi, Ankita Shukla, Yuan Wang, Hyunglae Lee, Matthew P. Buman, Pavan Turaga

Abstract: Deep learning methods have achieved a lot of success in various applications involving converting wearable sensor data to actionable health insights. A common application areas is activity recognition, where deep-learning methods still suffer from limitations such as sensitivity to signal quality, sensor characteristic variations, and variability between subjects. To mitigate these issues, robust… ▽ More Deep learning methods have achieved a lot of success in various applications involving converting wearable sensor data to actionable health insights. A common application areas is activity recognition, where deep-learning methods still suffer from limitations such as sensitivity to signal quality, sensor characteristic variations, and variability between subjects. To mitigate these issues, robust features obtained by topological data analysis (TDA) have been suggested as a potential solution. However, there are two significant obstacles to using topological features in deep learning: (1) large computational load to extract topological features using TDA, and (2) different signal representations obtained from deep learning and TDA which makes fusion difficult. In this paper, to enable integration of the strengths of topological methods in deep-learning for time-series data, we propose to use two teacher networks, one trained on the raw time-series data, and another trained on persistence images generated by TDA methods. The distilled student model utilizes only the raw time-series data at test-time. This approach addresses both issues. The use of KD with multiple teachers utilizes complementary information, and results in a compact model with strong supervisory features and an integrated richer representation. To assimilate desirable information from different modalities, we design new constraints, including orthogonality imposed on feature correlation maps for improving feature expressiveness and allowing the student to easily learn from the teacher. Also, we apply an annealing strategy in KD for fast saturation and better accommodation from different features, while the knowledge gap between the teachers and student is reduced. Finally, a robust student model is distilled, which uses only the time-series data as an input, while implicitly preserving topological features. △ Less

Submitted 7 July, 2024; originally announced July 2024.

Comments: Engineering Applications of Artificial Intelligence 130, 107719

Journal ref: Engineering Applications of Artificial Intelligence, 130, 107719 (2024)

Stark: Social Long-Term Multi-Modal Conversation with Persona Commonsense Knowledge

Authors: Young-Jun Lee, Dokyong Lee, Junyoung Youn, Kyeongjin Oh, Byungsoo Ko, Jonghwan Hyeon, Ho-Jin Choi

Abstract: Humans share a wide variety of images related to their personal experiences within conversations via instant messaging tools. However, existing works focus on (1) image-sharing behavior in singular sessions, leading to limited long-term social interaction, and (2) a lack of personalized image-sharing behavior. In this work, we introduce Stark, a large-scale long-term multi-modal conversation datas… ▽ More Humans share a wide variety of images related to their personal experiences within conversations via instant messaging tools. However, existing works focus on (1) image-sharing behavior in singular sessions, leading to limited long-term social interaction, and (2) a lack of personalized image-sharing behavior. In this work, we introduce Stark, a large-scale long-term multi-modal conversation dataset that covers a wide range of social personas in a multi-modality format, time intervals, and images. To construct Stark automatically, we propose a novel multi-modal contextualization framework, Mcu, that generates long-term multi-modal dialogue distilled from ChatGPT and our proposed Plan-and-Execute image aligner. Using our Stark, we train a multi-modal conversation model, Ultron 7B, which demonstrates impressive visual imagination ability. Furthermore, we demonstrate the effectiveness of our dataset in human evaluation. We make our source code and dataset publicly available. △ Less

Submitted 4 July, 2024; originally announced July 2024.

Comments: Project website: https://stark-dataset.github.io

Enhancing Accuracy and Parameter-Efficiency of Neural Representations for Network Parameterization

Authors: Hongjun Choi, Jayaraman J. Thiagarajan, Ruben Glatt, Shusen Liu

Abstract: In this work, we investigate the fundamental trade-off regarding accuracy and parameter efficiency in the parameterization of neural network weights using predictor networks. We present a surprising finding that, when recovering the original model accuracy is the sole objective, it can be achieved effectively through the weight reconstruction objective alone. Additionally, we explore the underlyin… ▽ More In this work, we investigate the fundamental trade-off regarding accuracy and parameter efficiency in the parameterization of neural network weights using predictor networks. We present a surprising finding that, when recovering the original model accuracy is the sole objective, it can be achieved effectively through the weight reconstruction objective alone. Additionally, we explore the underlying factors for improving weight reconstruction under parameter-efficiency constraints, and propose a novel training scheme that decouples the reconstruction objective from auxiliary objectives such as knowledge distillation that leads to significant improvements compared to state-of-the-art approaches. Finally, these results pave way for more practical scenarios, where one needs to achieve improvements on both model accuracy and predictor network parameter-efficiency simultaneously. △ Less

Submitted 29 June, 2024; originally announced July 2024.

LiCS: Navigation using Learned-imitation on Cluttered Space

Authors: Joshua Julian Damanik, Jae-Won Jung, Chala Adane Deresa, Han-Lim Choi

Abstract: In this letter, we propose a robust and fast navigation system in a narrow indoor environment for UGV (Unmanned Ground Vehicle) using 2D LiDAR and odometry. We used behavior cloning with Transformer neural network to learn the optimization-based baseline algorithm. We inject Gaussian noise during expert demonstration to increase the robustness of learned policy. We evaluate the performance of LiCS… ▽ More In this letter, we propose a robust and fast navigation system in a narrow indoor environment for UGV (Unmanned Ground Vehicle) using 2D LiDAR and odometry. We used behavior cloning with Transformer neural network to learn the optimization-based baseline algorithm. We inject Gaussian noise during expert demonstration to increase the robustness of learned policy. We evaluate the performance of LiCS using both simulation and hardware experiments. It outperforms all other baselines in terms of navigation performance and can maintain its robust performance even on highly cluttered environments. During the hardware experiments, LiCS can maintain safe navigation at maximum speed of $1.5\ m/s$. △ Less

Submitted 21 June, 2024; originally announced June 2024.

Comments: 6 pages, 4 figures. This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible

GrowOVER: How Can LLMs Adapt to Growing Real-World Knowledge?

Authors: Dayoon Ko, Jinyoung Kim, Hahyeon Choi, Gunhee Kim

Abstract: In the real world, knowledge is constantly evolving, which can render existing knowledge-based datasets outdated. This unreliability highlights the critical need for continuous updates to ensure both accuracy and relevance in knowledge-intensive tasks. To address this, we propose GrowOVER-QA and GrowOVER-Dialogue, dynamic open-domain QA and dialogue benchmarks that undergo a continuous cycle of up… ▽ More In the real world, knowledge is constantly evolving, which can render existing knowledge-based datasets outdated. This unreliability highlights the critical need for continuous updates to ensure both accuracy and relevance in knowledge-intensive tasks. To address this, we propose GrowOVER-QA and GrowOVER-Dialogue, dynamic open-domain QA and dialogue benchmarks that undergo a continuous cycle of updates, keeping pace with the rapid evolution of knowledge. Our research indicates that retrieval-augmented language models (RaLMs) struggle with knowledge that has not been trained on or recently updated. Consequently, we introduce a novel retrieval-interactive language model framework, where the language model evaluates and reflects on its answers for further re-retrieval. Our exhaustive experiments demonstrate that our training-free framework significantly improves upon existing methods, performing comparably to or even surpassing continuously trained language models. △ Less

Submitted 8 June, 2024; originally announced June 2024.

Comments: ACL 2024 Main

BWS: Best Window Selection Based on Sample Scores for Data Pruning across Broad Ranges

Authors: Hoyong Choi, Nohyun Ki, Hye Won Chung

Abstract: Data subset selection aims to find a smaller yet informative subset of a large dataset that can approximate the full-dataset training, addressing challenges associated with training neural networks on large-scale datasets. However, existing methods tend to specialize in either high or low selection ratio regimes, lacking a universal approach that consistently achieves competitive performance acros… ▽ More Data subset selection aims to find a smaller yet informative subset of a large dataset that can approximate the full-dataset training, addressing challenges associated with training neural networks on large-scale datasets. However, existing methods tend to specialize in either high or low selection ratio regimes, lacking a universal approach that consistently achieves competitive performance across a broad range of selection ratios. We introduce a universal and efficient data subset selection method, Best Window Selection (BWS), by proposing a method to choose the best window subset from samples ordered based on their difficulty scores. This approach offers flexibility by allowing the choice of window intervals that span from easy to difficult samples. Furthermore, we provide an efficient mechanism for selecting the best window subset by evaluating its quality using kernel ridge regression. Our experimental results demonstrate the superior performance of BWS compared to other baselines across a broad range of selection ratios over datasets, including CIFAR-10/100 and ImageNet, and the scenarios involving training from random initialization or fine-tuning of pre-trained models. △ Less

Submitted 5 June, 2024; originally announced June 2024.

Comments: ICML 2024

Quantum Circuit Switching with One-Way Repeaters in Star Networks

Authors: Álvaro G. Iñesta, Hyeongrak Choi, Dirk Englund, Stephanie Wehner

Abstract: Distributing quantum states reliably among distant locations is a key challenge in the field of quantum networks. One-way quantum networks address this by using one-way communication and quantum error correction. Here, we analyze quantum circuit switching as a protocol to distribute quantum states in one-way quantum networks. In quantum circuit switching, pairs of users can request the delivery of… ▽ More Distributing quantum states reliably among distant locations is a key challenge in the field of quantum networks. One-way quantum networks address this by using one-way communication and quantum error correction. Here, we analyze quantum circuit switching as a protocol to distribute quantum states in one-way quantum networks. In quantum circuit switching, pairs of users can request the delivery of multiple quantum states from one user to the other. After waiting for approval from the network, the states can be distributed either sequentially, forwarding one at a time along a path of quantum repeaters, or in parallel, sending batches of quantum states from repeater to repeater. Since repeaters can only forward a finite number of quantum states at a time, a pivotal question arises: is it advantageous to send them sequentially (allowing for multiple requests simultaneously) or in parallel (reducing processing time but handling only one request at a time)? We compare both approaches in a quantum network with a star topology. Using tools from queuing theory, we show that requests are met at a higher rate when packets are distributed in parallel, although sequential distribution can generally provide service to a larger number of users simultaneously. We also show that using a large number of quantum repeaters to combat channel losses limits the maximum distance between users, as each repeater introduces additional processing delays. These findings provide insight into the design of protocols for distributing quantum states in one-way quantum networks. △ Less

Submitted 29 May, 2024; originally announced May 2024.

Comments: Main text: 9 pages, 5 figures. Appendices: 14 pages, 8 figures

PICLe: Eliciting Diverse Behaviors from Large Language Models with Persona In-Context Learning

Authors: Hyeong Kyu Choi, Yixuan Li

Abstract: Large Language Models (LLMs) are trained on massive text corpora, which are encoded with diverse personality traits. This triggers an interesting goal of eliciting a desired personality trait from the LLM, and probing its behavioral preferences. Accordingly, we formalize the persona elicitation task, aiming to customize LLM behaviors to align with a target persona. We present Persona In-Context Le… ▽ More Large Language Models (LLMs) are trained on massive text corpora, which are encoded with diverse personality traits. This triggers an interesting goal of eliciting a desired personality trait from the LLM, and probing its behavioral preferences. Accordingly, we formalize the persona elicitation task, aiming to customize LLM behaviors to align with a target persona. We present Persona In-Context Learning (PICLe), a novel persona elicitation framework grounded in Bayesian inference. At the core, PICLe introduces a new ICL example selection criterion based on likelihood ratio, which is designed to optimally guide the model in eliciting a specific target persona. We demonstrate the effectiveness of PICLe through extensive comparisons against baseline methods across three contemporary LLMs. Code is available at https://github.com/deeplearning-wisc/picle. △ Less

Submitted 14 May, 2024; v1 submitted 3 May, 2024; originally announced May 2024.

Comments: ICML 2024

COPAL: Continual Pruning in Large Language Generative Models

Authors: Srikanth Malla, Joon Hee Choi, Chiho Choi

Abstract: Adapting pre-trained large language models to different domains in natural language processing requires two key considerations: high computational demands and model's inability to continual adaptation. To simultaneously address both issues, this paper presents COPAL (COntinual Pruning in Adaptive Language settings), an algorithm developed for pruning large language generative models under a contin… ▽ More Adapting pre-trained large language models to different domains in natural language processing requires two key considerations: high computational demands and model's inability to continual adaptation. To simultaneously address both issues, this paper presents COPAL (COntinual Pruning in Adaptive Language settings), an algorithm developed for pruning large language generative models under a continual model adaptation setting. While avoiding resource-heavy finetuning or retraining, our pruning process is guided by the proposed sensitivity analysis. The sensitivity effectively measures model's ability to withstand perturbations introduced by the new dataset and finds model's weights that are relevant for all encountered datasets. As a result, COPAL allows seamless model adaptation to new domains while enhancing the resource efficiency. Our empirical evaluation on a various size of LLMs show that COPAL outperforms baseline models, demonstrating its efficacy in efficiency and adaptability. △ Less

Submitted 14 June, 2024; v1 submitted 2 May, 2024; originally announced May 2024.

Comments: ICML2024

Distilling Privileged Information for Dubins Traveling Salesman Problems with Neighborhoods

Authors: Min Kyu Shin, Su-Jeong Park, Seung-Keol Ryu, Heeyeon Kim, Han-Lim Choi

Abstract: This paper presents a novel learning approach for Dubins Traveling Salesman Problems(DTSP) with Neighborhood (DTSPN) to quickly produce a tour of a non-holonomic vehicle passing through neighborhoods of given task points. The method involves two learning phases: initially, a model-free reinforcement learning approach leverages privileged information to distill knowledge from expert trajectories ge… ▽ More This paper presents a novel learning approach for Dubins Traveling Salesman Problems(DTSP) with Neighborhood (DTSPN) to quickly produce a tour of a non-holonomic vehicle passing through neighborhoods of given task points. The method involves two learning phases: initially, a model-free reinforcement learning approach leverages privileged information to distill knowledge from expert trajectories generated by the LinKernighan heuristic (LKH) algorithm. Subsequently, a supervised learning phase trains an adaptation network to solve problems independently of privileged information. Before the first learning phase, a parameter initialization technique using the demonstration data was also devised to enhance training efficiency. The proposed learning method produces a solution about 50 times faster than LKH and substantially outperforms other imitation learning and RL with demonstration schemes, most of which fail to sense all the task points. △ Less

Submitted 25 April, 2024; originally announced April 2024.

Comments: 7 pages, 4 figures, double blind under review

Pegasus-v1 Technical Report

Authors: Raehyuk Jung, Hyojun Go, Jaehyuk Yi, Jiho Jang, Daniel Kim, Jay Suh, Aiden Lee, Cooper Han, Jae Lee, Jeff Kim, Jin-Young Kim, Junwan Kim, Kyle Park, Lucas Lee, Mars Ha, Minjoon Seo, Abraham Jo, Ed Park, Hassan Kianinejad, SJ Kim, Tony Moon, Wade Jeong, Andrei Popescu, Esther Kim, EK Yoon , et al. (19 additional authors not shown)

Abstract: This technical report introduces Pegasus-1, a multimodal language model specialized in video content understanding and interaction through natural language. Pegasus-1 is designed to address the unique challenges posed by video data, such as interpreting spatiotemporal information, to offer nuanced video content comprehension across various lengths. This technical report overviews Pegasus-1's archi… ▽ More This technical report introduces Pegasus-1, a multimodal language model specialized in video content understanding and interaction through natural language. Pegasus-1 is designed to address the unique challenges posed by video data, such as interpreting spatiotemporal information, to offer nuanced video content comprehension across various lengths. This technical report overviews Pegasus-1's architecture, training strategies, and its performance in benchmarks on video conversation, zero-shot video question answering, and video summarization. We also explore qualitative characteristics of Pegasus-1 , demonstrating its capabilities as well as its limitations, in order to provide readers a balanced view of its current state and its future direction. △ Less

Submitted 22 April, 2024; originally announced April 2024.

A rate-distortion framework for MCMC algorithms: geometry and factorization of multivariate Markov chains

Authors: Michael C. H. Choi, Youjia Wang, Geoffrey Wolfer

Abstract: We introduce a framework rooted in a rate distortion problem for Markov chains, and show how a suite of commonly used Markov Chain Monte Carlo (MCMC) algorithms are specific instances within it, where the target stationary distribution is controlled by the distortion function. Our approach offers a unified variational view on the optimality of algorithms such as Metropolis-Hastings, Glauber dynami… ▽ More We introduce a framework rooted in a rate distortion problem for Markov chains, and show how a suite of commonly used Markov Chain Monte Carlo (MCMC) algorithms are specific instances within it, where the target stationary distribution is controlled by the distortion function. Our approach offers a unified variational view on the optimality of algorithms such as Metropolis-Hastings, Glauber dynamics, the swapping algorithm and Feynman-Kac path models. Along the way, we analyze factorizability and geometry of multivariate Markov chains. Specifically, we demonstrate that induced chains on factors of a product space can be regarded as information projections with respect to a particular divergence. This perspective yields Han--Shearer type inequalities for Markov chains as well as applications in the context of large deviations and mixing time comparison. △ Less

Submitted 18 April, 2024; originally announced April 2024.

Comments: 63 pages, 6 figures

MSC Class: 60F10; 60J10; 60J22; 94A15; 94A17

Emerging Property of Masked Token for Effective Pre-training

Authors: Hyesong Choi, Hunsang Lee, Seyoung Joung, Hyejin Park, Jiyeong Kim, Dongbo Min

Abstract: Driven by the success of Masked Language Modeling (MLM), the realm of self-supervised learning for computer vision has been invigorated by the central role of Masked Image Modeling (MIM) in driving recent breakthroughs. Notwithstanding the achievements of MIM across various downstream tasks, its overall efficiency is occasionally hampered by the lengthy duration of the pre-training phase. This pap… ▽ More Driven by the success of Masked Language Modeling (MLM), the realm of self-supervised learning for computer vision has been invigorated by the central role of Masked Image Modeling (MIM) in driving recent breakthroughs. Notwithstanding the achievements of MIM across various downstream tasks, its overall efficiency is occasionally hampered by the lengthy duration of the pre-training phase. This paper presents a perspective that the optimization of masked tokens as a means of addressing the prevailing issue. Initially, we delve into an exploration of the inherent properties that a masked token ought to possess. Within the properties, we principally dedicated to articulating and emphasizing the `data singularity' attribute inherent in masked tokens. Through a comprehensive analysis of the heterogeneity between masked tokens and visible tokens within pre-trained models, we propose a novel approach termed masked token optimization (MTO), specifically designed to improve model efficiency through weight recalibration and the enhancement of the key property of masked tokens. The proposed method serves as an adaptable solution that seamlessly integrates into any MIM approach that leverages masked tokens. As a result, MTO achieves a considerable improvement in pre-training efficiency, resulting in an approximately 50% reduction in pre-training epochs required to attain converged performance of the recent approaches. △ Less

Submitted 12 April, 2024; originally announced April 2024.

Salience-Based Adaptive Masking: Revisiting Token Dynamics for Enhanced Pre-training

Authors: Hyesong Choi, Hyejin Park, Kwang Moo Yi, Sungmin Cha, Dongbo Min

Abstract: In this paper, we introduce Saliency-Based Adaptive Masking (SBAM), a novel and cost-effective approach that significantly enhances the pre-training performance of Masked Image Modeling (MIM) approaches by prioritizing token salience. Our method provides robustness against variations in masking ratios, effectively mitigating the performance instability issues common in existing methods. This relax… ▽ More In this paper, we introduce Saliency-Based Adaptive Masking (SBAM), a novel and cost-effective approach that significantly enhances the pre-training performance of Masked Image Modeling (MIM) approaches by prioritizing token salience. Our method provides robustness against variations in masking ratios, effectively mitigating the performance instability issues common in existing methods. This relaxes the sensitivity of MIM-based pre-training to masking ratios, which in turn allows us to propose an adaptive strategy for `tailored' masking ratios for each data sample, which no existing method can provide. Toward this goal, we propose an Adaptive Masking Ratio (AMR) strategy that dynamically adjusts the proportion of masking for the unique content of each image based on token salience. We show that our method significantly improves over the state-of-the-art in mask-based pre-training on the ImageNet-1K dataset. △ Less

Submitted 12 April, 2024; originally announced April 2024.

PAAM: A Framework for Coordinated and Priority-Driven Accelerator Management in ROS 2

Authors: Daniel Enright, Yecheng Xiang, Hyunjong Choi, Hyoseung Kim

Abstract: This paper proposes a Priority-driven Accelerator Access Management (PAAM) framework for multi-process robotic applications built on top of the Robot Operating System (ROS) 2 middleware platform. The framework addresses the issue of predictable execution of time- and safety-critical callback chains that require hardware accelerators such as GPUs and TPUs. PAAM provides a standalone ROS executor th… ▽ More This paper proposes a Priority-driven Accelerator Access Management (PAAM) framework for multi-process robotic applications built on top of the Robot Operating System (ROS) 2 middleware platform. The framework addresses the issue of predictable execution of time- and safety-critical callback chains that require hardware accelerators such as GPUs and TPUs. PAAM provides a standalone ROS executor that acts as an accelerator resource server, arbitrating accelerator access requests from all other callbacks at the application layer. This approach enables coordinated and priority-driven accelerator access management in multi-process robotic systems. The framework design is directly applicable to all types of accelerators and enables granular control over how specific chains access accelerators, making it possible to achieve predictable real-time support for accelerators used by safety-critical callback chains without making changes to underlying accelerator device drivers. The paper shows that PAAM also offers a theoretical analysis that can upper bound the worst-case response time of safety-critical callback chains that necessitate accelerator access. This paper also demonstrates that complex robotic systems with extensive accelerator usage that are integrated with PAAM may achieve up to a 91\% reduction in end-to-end response time of their critical callback chains. △ Less

Submitted 9 April, 2024; originally announced April 2024.

Comments: 14 Pages, 14 Figures

Enhancing Clinical Efficiency through LLM: Discharge Note Generation for Cardiac Patients

Authors: HyoJe Jung, Yunha Kim, Heejung Choi, Hyeram Seo, Minkyoung Kim, JiYe Han, Gaeun Kee, Seohyun Park, Soyoung Ko, Byeolhee Kim, Suyeon Kim, Tae Joon Jun, Young-Hak Kim

Abstract: Medical documentation, including discharge notes, is crucial for ensuring patient care quality, continuity, and effective medical communication. However, the manual creation of these documents is not only time-consuming but also prone to inconsistencies and potential errors. The automation of this documentation process using artificial intelligence (AI) represents a promising area of innovation in… ▽ More Medical documentation, including discharge notes, is crucial for ensuring patient care quality, continuity, and effective medical communication. However, the manual creation of these documents is not only time-consuming but also prone to inconsistencies and potential errors. The automation of this documentation process using artificial intelligence (AI) represents a promising area of innovation in healthcare. This study directly addresses the inefficiencies and inaccuracies in creating discharge notes manually, particularly for cardiac patients, by employing AI techniques, specifically large language model (LLM). Utilizing a substantial dataset from a cardiology center, encompassing wide-ranging medical records and physician assessments, our research evaluates the capability of LLM to enhance the documentation process. Among the various models assessed, Mistral-7B distinguished itself by accurately generating discharge notes that significantly improve both documentation efficiency and the continuity of care for patients. These notes underwent rigorous qualitative evaluation by medical expert, receiving high marks for their clinical relevance, completeness, readability, and contribution to informed decision-making and care planning. Coupled with quantitative analyses, these results confirm Mistral-7B's efficacy in distilling complex medical information into concise, coherent summaries. Overall, our findings illuminate the considerable promise of specialized LLM, such as Mistral-7B, in refining healthcare documentation workflows and advancing patient care. This study lays the groundwork for further integrating advanced AI technologies in healthcare, demonstrating their potential to revolutionize patient documentation and support better care outcomes. △ Less

Submitted 7 April, 2024; originally announced April 2024.

Comments: 10 pages, 1 figure, 3 tables, conference

HyperCLOVA X Technical Report

Authors: Kang Min Yoo, Jaegeun Han, Sookyo In, Heewon Jeon, Jisu Jeong, Jaewook Kang, Hyunwook Kim, Kyung-Min Kim, Munhyong Kim, Sungju Kim, Donghyun Kwak, Hanock Kwak, Se Jung Kwon, Bado Lee, Dongsoo Lee, Gichang Lee, Jooho Lee, Baeseong Park, Seongjin Shin, Joonsang Yu, Seolki Baek, Sumin Byeon, Eungsup Cho, Dooseok Choe, Jeesung Han , et al. (371 additional authors not shown)

Abstract: We introduce HyperCLOVA X, a family of large language models (LLMs) tailored to the Korean language and culture, along with competitive capabilities in English, math, and coding. HyperCLOVA X was trained on a balanced mix of Korean, English, and code data, followed by instruction-tuning with high-quality human-annotated datasets while abiding by strict safety guidelines reflecting our commitment t… ▽ More We introduce HyperCLOVA X, a family of large language models (LLMs) tailored to the Korean language and culture, along with competitive capabilities in English, math, and coding. HyperCLOVA X was trained on a balanced mix of Korean, English, and code data, followed by instruction-tuning with high-quality human-annotated datasets while abiding by strict safety guidelines reflecting our commitment to responsible AI. The model is evaluated across various benchmarks, including comprehensive reasoning, knowledge, commonsense, factuality, coding, math, chatting, instruction-following, and harmlessness, in both Korean and English. HyperCLOVA X exhibits strong reasoning capabilities in Korean backed by a deep understanding of the language and cultural nuances. Further analysis of the inherent bilingual nature and its extension to multilingualism highlights the model's cross-lingual proficiency and strong generalization ability to untargeted languages, including machine translation between several language pairs and cross-lingual inference tasks. We believe that HyperCLOVA X can provide helpful guidance for regions or countries in developing their sovereign LLMs. △ Less

Submitted 13 April, 2024; v1 submitted 2 April, 2024; originally announced April 2024.

Comments: 44 pages; updated authors list and fixed author names

Advancing AI with Integrity: Ethical Challenges and Solutions in Neural Machine Translation

Authors: Richard Kimera, Yun-Seon Kim, Heeyoul Choi

Abstract: This paper addresses the ethical challenges of Artificial Intelligence in Neural Machine Translation (NMT) systems, emphasizing the imperative for developers to ensure fairness and cultural sensitivity. We investigate the ethical competence of AI models in NMT, examining the Ethical considerations at each stage of NMT development, including data handling, privacy, data ownership, and consent. We i… ▽ More This paper addresses the ethical challenges of Artificial Intelligence in Neural Machine Translation (NMT) systems, emphasizing the imperative for developers to ensure fairness and cultural sensitivity. We investigate the ethical competence of AI models in NMT, examining the Ethical considerations at each stage of NMT development, including data handling, privacy, data ownership, and consent. We identify and address ethical issues through empirical studies. These include employing Transformer models for Luganda-English translations and enhancing efficiency with sentence mini-batching. And complementary studies that refine data labeling techniques and fine-tune BERT and Longformer models for analyzing Luganda and English social media content. Our second approach is a literature review from databases such as Google Scholar and platforms like GitHub. Additionally, the paper probes the distribution of responsibility between AI systems and humans, underscoring the essential role of human oversight in upholding NMT ethical standards. Incorporating a biblical perspective, we discuss the societal impact of NMT and the broader ethical responsibilities of developers, positing them as stewards accountable for the societal repercussions of their creations. △ Less

Submitted 1 April, 2024; originally announced April 2024.

Comments: 11 pages

Creating Aesthetic Sonifications on the Web with SIREN

Authors: Tristan Peng, Hongchan Choi, Jonathan Berger

Abstract: SIREN is a flexible, extensible, and customizable web-based general-purpose interface for auditory data display (sonification). Designed as a digital audio workstation for sonification, synthesizers written in JavaScript using the Web Audio API facilitate intuitive mapping of data to auditory parameters for a wide range of purposes. This paper explores the breadth of sound synthesis techniques s… ▽ More SIREN is a flexible, extensible, and customizable web-based general-purpose interface for auditory data display (sonification). Designed as a digital audio workstation for sonification, synthesizers written in JavaScript using the Web Audio API facilitate intuitive mapping of data to auditory parameters for a wide range of purposes. This paper explores the breadth of sound synthesis techniques supported by SIREN, and details the structure and definition of a SIREN synthesizer module. The paper proposes further development that will increase SIREN's utility. △ Less

Submitted 28 March, 2024; originally announced March 2024.

Comments: 7 pages, 1 figure, 5 listings, submitted to the Web Audio Conference 2024

WMMSE-Based Rate Maximization for RIS-Assisted MU-MIMO Systems

Authors: Hyuckjin Choi, A. Lee Swindlehurst, Junil Choi

Abstract: Reconfigurable intelligent surface (RIS) technology, given its ability to favorably modify wireless communication environments, will play a pivotal role in the evolution of future communication systems. This paper proposes rate maximization techniques for both single-user and multiuser MIMO systems, based on the well-known weighted minimum mean square error (WMMSE) criterion. Using a suitable weig… ▽ More Reconfigurable intelligent surface (RIS) technology, given its ability to favorably modify wireless communication environments, will play a pivotal role in the evolution of future communication systems. This paper proposes rate maximization techniques for both single-user and multiuser MIMO systems, based on the well-known weighted minimum mean square error (WMMSE) criterion. Using a suitable weight matrix, the WMMSE algorithm tackles an equivalent weighted mean square error (WMSE) minimization problem to achieve the sum-rate maximization. By considering a more practical RIS system model that employs a tensor-based representation enforced by the electromagnetic behavior exhibited by the RIS panel, we detail both the sum-rate maximizing and WMSE minimizing strategies for RIS phase shift optimization by deriving the closed-form gradient of the WMSE and the sum-rate with respect to the RIS phase shift vector. Our simulations reveal that the proposed rate maximization technique, rooted in the WMMSE algorithm, exhibits superior performance when compared to other benchmarks. △ Less

Submitted 19 March, 2024; originally announced March 2024.

A Deep Reinforcement Learning Approach for Autonomous Reconfigurable Intelligent Surfaces

Authors: Hyuckjin Choi, Ly V. Nguyen, Junil Choi, A. Lee Swindlehurst

Abstract: A reconfigurable intelligent surface (RIS) is a prospective wireless technology that enhances wireless channel quality. An RIS is often equipped with passive array of elements and provides cost and power-efficient solutions for coverage extension of wireless communication systems. Without any radio frequency (RF) chains or computing resources, however, the RIS requires control information to be se… ▽ More A reconfigurable intelligent surface (RIS) is a prospective wireless technology that enhances wireless channel quality. An RIS is often equipped with passive array of elements and provides cost and power-efficient solutions for coverage extension of wireless communication systems. Without any radio frequency (RF) chains or computing resources, however, the RIS requires control information to be sent to it from an external unit, e.g., a base station (BS). The control information can be delivered by wired or wireless channels, and the BS must be aware of the RIS and the RIS-related channel conditions in order to effectively configure its behavior. Recent works have introduced hybrid RIS structures possessing a few active elements that can sense and digitally process received data. Here, we propose the operation of an entirely autonomous RIS that operates without a control link between the RIS and BS. Using a few sensing elements, the autonomous RIS employs a deep Q network (DQN) based on reinforcement learning in order to enhance the sum rate of the network. Our results illustrate the potential of deploying autonomous RISs in wireless networks with essentially no network overhead. △ Less

Submitted 19 March, 2024; v1 submitted 14 March, 2024; originally announced March 2024.

On the Feasibility of EEG-based Motor Intention Detection for Real-Time Robot Assistive Control

Authors: Ho Jin Choi, Satyajeet Das, Shaoting Peng, Ruzena Bajcsy, Nadia Figueroa

Abstract: This paper explores the feasibility of employing EEG-based intention detection for real-time robot assistive control. We focus on predicting and distinguishing motor intentions of left/right arm movements by presenting: i) an offline data collection and training pipeline, used to train a classifier for left/right motion intention prediction, and ii) an online real-time prediction pipeline leveragi… ▽ More This paper explores the feasibility of employing EEG-based intention detection for real-time robot assistive control. We focus on predicting and distinguishing motor intentions of left/right arm movements by presenting: i) an offline data collection and training pipeline, used to train a classifier for left/right motion intention prediction, and ii) an online real-time prediction pipeline leveraging the trained classifier and integrated with an assistive robot. Central to our approach is a rich feature representation composed of the tangent space projection of time-windowed sample covariance matrices from EEG filtered signals and derivatives; allowing for a simple SVM classifier to achieve unprecedented accuracy and real-time performance. In pre-recorded real-time settings (160 Hz), a peak accuracy of 86.88% is achieved, surpassing prior works. In robot-in-the-loop settings, our system successfully detects intended motion solely from EEG data with 70% accuracy, triggering a robot to execute an assistive task. We provide a comprehensive evaluation of the proposed classifier. △ Less

Submitted 12 March, 2024; originally announced March 2024.

Style Blind Domain Generalized Semantic Segmentation via Covariance Alignment and Semantic Consistence Contrastive Learning

Authors: Woo-Jin Ahn, Geun-Yeong Yang, Hyun-Duck Choi, Myo-Taeg Lim

Abstract: Deep learning models for semantic segmentation often experience performance degradation when deployed to unseen target domains unidentified during the training phase. This is mainly due to variations in image texture (\ie style) from different data sources. To tackle this challenge, existing domain generalized semantic segmentation (DGSS) methods attempt to remove style variations from the feature… ▽ More Deep learning models for semantic segmentation often experience performance degradation when deployed to unseen target domains unidentified during the training phase. This is mainly due to variations in image texture (\ie style) from different data sources. To tackle this challenge, existing domain generalized semantic segmentation (DGSS) methods attempt to remove style variations from the feature. However, these approaches struggle with the entanglement of style and content, which may lead to the unintentional removal of crucial content information, causing performance degradation. This study addresses this limitation by proposing BlindNet, a novel DGSS approach that blinds the style without external modules or datasets. The main idea behind our proposed approach is to alleviate the effect of style in the encoder whilst facilitating robust segmentation in the decoder. To achieve this, BlindNet comprises two key components: covariance alignment and semantic consistency contrastive learning. Specifically, the covariance alignment trains the encoder to uniformly recognize various styles and preserve the content information of the feature, rather than removing the style-sensitive factor. Meanwhile, semantic consistency contrastive learning enables the decoder to construct discriminative class embedding space and disentangles features that are vulnerable to misclassification. Through extensive experiments, our approach outperforms existing DGSS methods, exhibiting robustness and superior performance for semantic segmentation on unseen target domains. △ Less

Submitted 10 March, 2024; originally announced March 2024.

Comments: CVPR 2024

Improving Diffusion Models for Virtual Try-on

Authors: Yisol Choi, Sangkyung Kwak, Kyungmin Lee, Hyungwon Choi, Jinwoo Shin

Abstract: This paper considers image-based virtual try-on, which renders an image of a person wearing a curated garment, given a pair of images depicting the person and the garment, respectively. Previous works adapt existing exemplar-based inpainting diffusion models for virtual try-on to improve the naturalness of the generated visuals compared to other methods (e.g., GAN-based), but they fail to preserve… ▽ More This paper considers image-based virtual try-on, which renders an image of a person wearing a curated garment, given a pair of images depicting the person and the garment, respectively. Previous works adapt existing exemplar-based inpainting diffusion models for virtual try-on to improve the naturalness of the generated visuals compared to other methods (e.g., GAN-based), but they fail to preserve the identity of the garments. To overcome this limitation, we propose a novel diffusion model that improves garment fidelity and generates authentic virtual try-on images. Our method, coined IDM-VTON, uses two different modules to encode the semantics of garment image; given the base UNet of the diffusion model, 1) the high-level semantics extracted from a visual encoder are fused to the cross-attention layer, and then 2) the low-level features extracted from parallel UNet are fused to the self-attention layer. In addition, we provide detailed textual prompts for both garment and person images to enhance the authenticity of the generated visuals. Finally, we present a customization method using a pair of person-garment images, which significantly improves fidelity and authenticity. Our experimental results show that our method outperforms previous approaches (both diffusion-based and GAN-based) in preserving garment details and generating authentic virtual try-on images, both qualitatively and quantitatively. Furthermore, the proposed customization method demonstrates its effectiveness in a real-world scenario. More visualizations are available in our project page: https://idm-vton.github.io △ Less

Submitted 19 March, 2024; v1 submitted 8 March, 2024; originally announced March 2024.

NeuPIMs: NPU-PIM Heterogeneous Acceleration for Batched LLM Inferencing

Authors: Guseul Heo, Sangyeop Lee, Jaehong Cho, Hyunmin Choi, Sanghyeon Lee, Hyungkyu Ham, Gwangsun Kim, Divya Mahajan, Jongse Park

Abstract: Modern transformer-based Large Language Models (LLMs) are constructed with a series of decoder blocks. Each block comprises three key components: (1) QKV generation, (2) multi-head attention, and (3) feed-forward networks. In batched processing, QKV generation and feed-forward networks involve compute-intensive matrix-matrix multiplications (GEMM), while multi-head attention requires bandwidth-hea… ▽ More Modern transformer-based Large Language Models (LLMs) are constructed with a series of decoder blocks. Each block comprises three key components: (1) QKV generation, (2) multi-head attention, and (3) feed-forward networks. In batched processing, QKV generation and feed-forward networks involve compute-intensive matrix-matrix multiplications (GEMM), while multi-head attention requires bandwidth-heavy matrix-vector multiplications (GEMV). Machine learning accelerators like TPUs or NPUs are proficient in handling GEMM but are less efficient for GEMV computations. Conversely, Processing-in-Memory (PIM) technology is tailored for efficient GEMV computation, while it lacks the computational power to handle GEMM effectively. Inspired by this insight, we propose NeuPIMs, a heterogeneous acceleration system that jointly exploits a conventional GEMM-focused NPU and GEMV-optimized PIM devices. The main challenge in efficiently integrating NPU and PIM lies in enabling concurrent operations on both platforms, each addressing a specific kernel type. First, existing PIMs typically operate in a "blocked" mode, allowing only either NPU or PIM to be active at any given time. Second, the inherent dependencies between GEMM and GEMV in LLMs restrict their parallel processing. To tackle these challenges, NeuPIMs is equipped with dual row buffers in each bank, facilitating the simultaneous management of memory read/write operations and PIM commands. Further, NeuPIMs employs a runtime sub-batch interleaving technique to maximize concurrent execution, leveraging batch parallelism to allow two independent sub-batches to be pipelined within a single NeuPIMs device. Our evaluation demonstrates that compared to GPU-only, NPU-only, and a naïve NPU+PIM integrated acceleration approaches, NeuPIMs achieves 3$\times$, 2.4$\times$ and 1.6$\times$ throughput improvement, respectively. △ Less

Submitted 29 March, 2024; v1 submitted 1 March, 2024; originally announced March 2024.

Comments: 16 pages, 15 figures

Journal ref: ASPLOS 2024

Variable-Rate Learned Image Compression with Multi-Objective Optimization and Quantization-Reconstruction Offsets

Authors: Fatih Kamisli, Fabien Racape, Hyomin Choi

Abstract: Achieving successful variable bitrate compression with computationally simple algorithms from a single end-to-end learned image or video compression model remains a challenge. Many approaches have been proposed, including conditional auto-encoders, channel-adaptive gains for the latent tensor or uniformly quantizing all elements of the latent tensor. This paper follows the traditional approach to… ▽ More Achieving successful variable bitrate compression with computationally simple algorithms from a single end-to-end learned image or video compression model remains a challenge. Many approaches have been proposed, including conditional auto-encoders, channel-adaptive gains for the latent tensor or uniformly quantizing all elements of the latent tensor. This paper follows the traditional approach to vary a single quantization step size to perform uniform quantization of all latent tensor elements. However, three modifications are proposed to improve the variable rate compression performance. First, multi objective optimization is used for (post) training. Second, a quantization-reconstruction offset is introduced into the quantization operation. Third, variable rate quantization is used also for the hyper latent. All these modifications can be made on a pre-trained single-rate compression model by performing post training. The algorithms are implemented into three well-known image compression models and the achieved variable rate compression results indicate negligible or minimal compression performance loss compared to training multiple models. (Codes will be shared at https://github.com/InterDigitalInc/CompressAI) △ Less

Submitted 29 February, 2024; originally announced February 2024.

Comments: Accepted as a paper at DCC 2024

Objective and Interpretable Breast Cosmesis Evaluation with Attention Guided Denoising Diffusion Anomaly Detection Model

Authors: Sangjoon Park, Yong Bae Kim, Jee Suk Chang, Seo Hee Choi, Hyungjin Chung, Ik Jae Lee, Hwa Kyung Byun

Abstract: As advancements in the field of breast cancer treatment continue to progress, the assessment of post-surgical cosmetic outcomes has gained increasing significance due to its substantial impact on patients' quality of life. However, evaluating breast cosmesis presents challenges due to the inherently subjective nature of expert labeling. In this study, we present a novel automated approach, Attenti… ▽ More As advancements in the field of breast cancer treatment continue to progress, the assessment of post-surgical cosmetic outcomes has gained increasing significance due to its substantial impact on patients' quality of life. However, evaluating breast cosmesis presents challenges due to the inherently subjective nature of expert labeling. In this study, we present a novel automated approach, Attention-Guided Denoising Diffusion Anomaly Detection (AG-DDAD), designed to assess breast cosmesis following surgery, addressing the limitations of conventional supervised learning and existing anomaly detection models. Our approach leverages the attention mechanism of the distillation with no label (DINO) self-supervised Vision Transformer (ViT) in combination with a diffusion model to achieve high-quality image reconstruction and precise transformation of discriminative regions. By training the diffusion model on unlabeled data predominantly with normal cosmesis, we adopt an unsupervised anomaly detection perspective to automatically score the cosmesis. Real-world data experiments demonstrate the effectiveness of our method, providing visually appealing representations and quantifiable scores for cosmesis evaluation. Compared to commonly used rule-based programs, our fully automated approach eliminates the need for manual annotations and offers objective evaluation. Moreover, our anomaly detection model exhibits state-of-the-art performance, surpassing existing models in accuracy. Going beyond the scope of breast cosmesis, our research represents a significant advancement in unsupervised anomaly detection within the medical domain, thereby paving the way for future investigations. △ Less

Submitted 28 February, 2024; originally announced February 2024.

Tactile-Informed Action Primitives Mitigate Jamming in Dense Clutter

Authors: Dane Brouwer, Joshua Citron, Hojung Choi, Marion Lepert, Michael Lin, Jeannette Bohg, Mark Cutkosky

Abstract: It is difficult for robots to retrieve objects in densely cluttered lateral access scenes with movable objects as jamming against adjacent objects and walls can inhibit progress. We propose the use of two action primitives -- burrowing and excavating -- that can fluidize the scene to un-jam obstacles and enable continued progress. Even when these primitives are implemented in an open loop manner a… ▽ More It is difficult for robots to retrieve objects in densely cluttered lateral access scenes with movable objects as jamming against adjacent objects and walls can inhibit progress. We propose the use of two action primitives -- burrowing and excavating -- that can fluidize the scene to un-jam obstacles and enable continued progress. Even when these primitives are implemented in an open loop manner at clock-driven intervals, we observe a decrease in the final distance to the target location. Furthermore, we combine the primitives into a closed loop hybrid control strategy using tactile and proprioceptive information to leverage the advantages of both primitives without being overly disruptive. In doing so, we achieve a 10-fold increase in success rate above the baseline control strategy and significantly improve completion times as compared to the primitives alone or a naive combination of them. △ Less

Submitted 14 February, 2024; originally announced February 2024.

Comments: Preprint of paper accepted to IEEE ICRA 2024

UniHENN: Designing More Versatile Homomorphic Encryption-based CNNs without im2col

Authors: Hyunmin Choi, Jihun Kim, Seungho Kim, Seonhye Park, Jeongyong Park, Wonbin Choi, Hyoungshick Kim

Abstract: Homomorphic encryption enables computations on encrypted data without decryption, which is crucial for privacy-preserving cloud services. However, deploying convolutional neural networks (CNNs) with homomorphic encryption encounters significant challenges, particularly in converting input data into a two-dimensional matrix for convolution, typically achieved using the im2col technique. While effic… ▽ More Homomorphic encryption enables computations on encrypted data without decryption, which is crucial for privacy-preserving cloud services. However, deploying convolutional neural networks (CNNs) with homomorphic encryption encounters significant challenges, particularly in converting input data into a two-dimensional matrix for convolution, typically achieved using the im2col technique. While efficient, this method limits the variety of deployable CNN models due to compatibility constraints with the encrypted data structure. UniHENN, a homomorphic encryption-based CNN architecture, eliminates the need for im2col, ensuring compatibility with a diverse range of CNN models using homomorphic encryption. Our experiments demonstrate that UniHENN surpasses the leading 2D CNN inference architecture, PyCrCNN, in inference time, as evidenced by its performance on the LeNet-1 dataset, where it averages 30.090 seconds--significantly faster than PyCrCNN's 794.064 seconds. Furthermore, UniHENN outperforms TenSEAL, which employs im2col, in processing concurrent images, an essential feature for high-demand cloud applications. The versatility of UniHENN is proven across various CNN architectures, including 1D and six different 2D CNNs, highlighting its flexibility and efficiency. These qualities establish UniHENN as a promising solution for privacy-preserving, cloud-based CNN services, addressing the increasing demand for scalable, secure, and efficient deep learning in cloud computing environments. △ Less

Submitted 5 February, 2024; originally announced February 2024.

Multi-Agent Based Transfer Learning for Data-Driven Air Traffic Applications

Authors: Chuhao Deng, Hong-Cheol Choi, Hyunsang Park, Inseok Hwang

Abstract: Research in developing data-driven models for Air Traffic Management (ATM) has gained a tremendous interest in recent years. However, data-driven models are known to have long training time and require large datasets to achieve good performance. To address the two issues, this paper proposes a Multi-Agent Bidirectional Encoder Representations from Transformers (MA-BERT) model that fully considers… ▽ More Research in developing data-driven models for Air Traffic Management (ATM) has gained a tremendous interest in recent years. However, data-driven models are known to have long training time and require large datasets to achieve good performance. To address the two issues, this paper proposes a Multi-Agent Bidirectional Encoder Representations from Transformers (MA-BERT) model that fully considers the multi-agent characteristic of the ATM system and learns air traffic controllers' decisions, and a pre-training and fine-tuning transfer learning framework. By pre-training the MA-BERT on a large dataset from a major airport and then fine-tuning it to other airports and specific air traffic applications, a large amount of the total training time can be saved. In addition, for newly adopted procedures and constructed airports where no historical data is available, this paper shows that the pre-trained MA-BERT can achieve high performance by updating regularly with little data. The proposed transfer learning framework and MA-BERT are tested with the automatic dependent surveillance-broadcast data recorded in 3 airports in South Korea in 2019. △ Less

Submitted 23 January, 2024; originally announced January 2024.

Comments: 12 pages, 8 figures, submitted for IEEE Transactions on Intelligent Transportation System

Enhanced Labeling Technique for Reddit Text and Fine-Tuned Longformer Models for Classifying Depression Severity in English and Luganda

Authors: Richard Kimera, Daniela N. Rim, Joseph Kirabira, Ubong Godwin Udomah, Heeyoul Choi

Abstract: Depression is a global burden and one of the most challenging mental health conditions to control. Experts can detect its severity early using the Beck Depression Inventory (BDI) questionnaire, administer appropriate medication to patients, and impede its progression. Due to the fear of potential stigmatization, many patients turn to social media platforms like Reddit for advice and assistance at… ▽ More Depression is a global burden and one of the most challenging mental health conditions to control. Experts can detect its severity early using the Beck Depression Inventory (BDI) questionnaire, administer appropriate medication to patients, and impede its progression. Due to the fear of potential stigmatization, many patients turn to social media platforms like Reddit for advice and assistance at various stages of their journey. This research extracts text from Reddit to facilitate the diagnostic process. It employs a proposed labeling approach to categorize the text and subsequently fine-tunes the Longformer model. The model's performance is compared against baseline models, including Naive Bayes, Random Forest, Support Vector Machines, and Gradient Boosting. Our findings reveal that the Longformer model outperforms the baseline models in both English (48%) and Luganda (45%) languages on a custom-made dataset. △ Less

Submitted 25 January, 2024; originally announced January 2024.

Comments: In IEEE Proceedings of the 14th International Conference on ICT Convergence (ICTC), Jeju, Korea, October 2023

Enhancing Wind Speed and Wind Power Forecasting Using Shape-Wise Feature Engineering: A Novel Approach for Improved Accuracy and Robustness

Authors: Mulomba Mukendi Christian, Yun Seon Kim, Hyebong Choi, Jaeyoung Lee, SongHee You

Abstract: Accurate prediction of wind speed and power is vital for enhancing the efficiency of wind energy systems. Numerous solutions have been implemented to date, demonstrating their potential to improve forecasting. Among these, deep learning is perceived as a revolutionary approach in the field. However, despite their effectiveness, the noise present in the collected data remains a significant challeng… ▽ More Accurate prediction of wind speed and power is vital for enhancing the efficiency of wind energy systems. Numerous solutions have been implemented to date, demonstrating their potential to improve forecasting. Among these, deep learning is perceived as a revolutionary approach in the field. However, despite their effectiveness, the noise present in the collected data remains a significant challenge. This noise has the potential to diminish the performance of these algorithms, leading to inaccurate predictions. In response to this, this study explores a novel feature engineering approach. This approach involves altering the data input shape in both Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) and Autoregressive models for various forecasting horizons. The results reveal substantial enhancements in model resilience against noise resulting from step increases in data. The approach could achieve an impressive 83% accuracy in predicting unseen data up to the 24th steps. Furthermore, this method consistently provides high accuracy for short, mid, and long-term forecasts, outperforming the performance of individual models. These findings pave the way for further research on noise reduction strategies at different forecasting horizons through shape-wise feature engineering. △ Less

Submitted 16 January, 2024; originally announced January 2024.

Journal ref: International Journal of Advanced Culture Technology Vol.11 No.4 393-405 (2023)

Temporal Analysis of World Disaster Risk:A Machine Learning Approach to Cluster Dynamics

Authors: Christian Mulomba Mukendi, Hyebong Choi

Abstract: he evaluation of the impact of actions undertaken is essential in management. This paper assesses the impact of efforts considered to mitigate risk and create safe environments on a global scale. We measure this impact by looking at the probability of improvement over a specific short period of time. Using the World Risk Index, we conduct a temporal analysis of global disaster risk dynamics from 2… ▽ More he evaluation of the impact of actions undertaken is essential in management. This paper assesses the impact of efforts considered to mitigate risk and create safe environments on a global scale. We measure this impact by looking at the probability of improvement over a specific short period of time. Using the World Risk Index, we conduct a temporal analysis of global disaster risk dynamics from 2011 to 2021. This temporal exploration through the lens of the World Risk Index provides insights into the complex dynamics of disaster risk. We found that, despite sustained efforts, the global landscape remains divided into two main clusters: high susceptibility and moderate susceptibility, regardless of geographical location. This clustering was achieved using a semi-supervised approach through the Label Spreading algorithm, with 98% accuracy. We also found that the prediction of clusters achieved through supervised learning on the period considered in this study (one, three, and five years) showed that the Logistic regression (almost 99% at each stage) performed better than other classifiers. This suggests that the current policies and mechanisms are not effective in helping countries move from a hazardous position to a safer one during the period considered. In fact, statistical projections using a scenario analysis indicate that there is only a 1% chance of such a shift occurring within a five-year timeframe. This sobering reality highlights the need for a paradigm shift. Traditional long-term disaster management strategies are not effective for countries that are highly vulnerable. Our findings indicate the need for an innovative approach that is tailored to the specific vulnerabilities of these nations. As the threat of vulnerability persists, our research calls for the development of new strategies that can effectively address the ongoing challenges of disaster risk management △ Less

Submitted 10 January, 2024; originally announced January 2024.

Comments: This is the conference proceeding of the ICTC 2023, to be published in IEEE conference proceedings

Report number: 979-8-3503-1327-7/23/$31.00 \c{opyright}2023 IEEE

Air Quality Forecasting Using Machine Learning: A Global perspective with Relevance to Low-Resource Settings

Authors: Mulomba Mukendi Christian, Hyebong Choi

Abstract: Air pollution stands as the fourth leading cause of death globally. While extensive research has been conducted in this domain, most approaches rely on large datasets when it comes to prediction. This limits their applicability in low-resource settings though more vulnerable. This study addresses this gap by proposing a novel machine learning approach for accurate air quality prediction using two… ▽ More Air pollution stands as the fourth leading cause of death globally. While extensive research has been conducted in this domain, most approaches rely on large datasets when it comes to prediction. This limits their applicability in low-resource settings though more vulnerable. This study addresses this gap by proposing a novel machine learning approach for accurate air quality prediction using two months of air quality data. By leveraging the World Weather Repository, the meteorological, air pollutant, and Air Quality Index features from 197 capital cities were considered to predict air quality for the next day. The evaluation of several machine learning models demonstrates the effectiveness of the Random Forest algorithm in generating reliable predictions, particularly when applied to classification rather than regression, approach which enhances the model's generalizability by 42%, achieving a cross-validation score of 0.38 for regression and 0.89 for classification. To instill confidence in the predictions, interpretable machine learning was considered. Finally, a cost estimation comparing the implementation of this solution in high-resource and low-resource settings is presented including a tentative of technology licensing business model. This research highlights the potential for resource-limited countries to independently predict air quality while awaiting larger datasets to further refine their predictions. △ Less

Submitted 9 January, 2024; originally announced January 2024.

Comments: 16 pages. This is a conference proceeding Presented at: SIBR 2024 (Seoul) Conference on Interdisciplinary Business and Economics Research, 5th-6th January 2024, Seoul, South Korea

Enhancing Acute Kidney Injury Prediction through Integration of Drug Features in Intensive Care Units

Authors: Gabriel D. M. Manalu, Mulomba Mukendi Christian, Songhee You, Hyebong Choi

Abstract: The relationship between acute kidney injury (AKI) prediction and nephrotoxic drugs, or drugs that adversely affect kidney function, is one that has yet to be explored in the critical care setting. One contributing factor to this gap in research is the limited investigation of drug modalities in the intensive care unit (ICU) context, due to the challenges of processing prescription data into the c… ▽ More The relationship between acute kidney injury (AKI) prediction and nephrotoxic drugs, or drugs that adversely affect kidney function, is one that has yet to be explored in the critical care setting. One contributing factor to this gap in research is the limited investigation of drug modalities in the intensive care unit (ICU) context, due to the challenges of processing prescription data into the corresponding drug representations and a lack in the comprehensive understanding of these drug representations. This study addresses this gap by proposing a novel approach that leverages patient prescription data as a modality to improve existing models for AKI prediction. We base our research on Electronic Health Record (EHR) data, extracting the relevant patient prescription information and converting it into the selected drug representation for our research, the extended-connectivity fingerprint (ECFP). Furthermore, we adopt a unique multimodal approach, developing machine learning models and 1D Convolutional Neural Networks (CNN) applied to clinical drug representations, establishing a procedure which has not been used by any previous studies predicting AKI. The findings showcase a notable improvement in AKI prediction through the integration of drug embeddings and other patient cohort features. By using drug features represented as ECFP molecular fingerprints along with common cohort features such as demographics and lab test values, we achieved a considerable improvement in model performance for the AKI prediction task over the baseline model which does not include the drug representations as features, indicating that our distinct approach enhances existing baseline techniques and highlights the relevance of drug data in predicting AKI in the ICU setting △ Less

Submitted 9 January, 2024; originally announced January 2024.

Comments: 9 pages, 2 tables

Journal ref: International Journal of Advanced Smart Convergence Vol.12 No.4 434- 442 (2023)

UFO: Unidentified Foreground Object Detection in 3D Point Cloud

Authors: Hyunjun Choi, Hawook Jeong, Jin Young Choi

Abstract: In this paper, we raise a new issue on Unidentified Foreground Object (UFO) detection in 3D point clouds, which is a crucial technology in autonomous driving in the wild. UFO detection is challenging in that existing 3D object detectors encounter extremely hard challenges in both 3D localization and Out-of-Distribution (OOD) detection. To tackle these challenges, we suggest a new UFO detection fra… ▽ More In this paper, we raise a new issue on Unidentified Foreground Object (UFO) detection in 3D point clouds, which is a crucial technology in autonomous driving in the wild. UFO detection is challenging in that existing 3D object detectors encounter extremely hard challenges in both 3D localization and Out-of-Distribution (OOD) detection. To tackle these challenges, we suggest a new UFO detection framework including three tasks: evaluation protocol, methodology, and benchmark. The evaluation includes a new approach to measure the performance on our goal, i.e. both localization and OOD detection of UFOs. The methodology includes practical techniques to enhance the performance of our goal. The benchmark is composed of the KITTI Misc benchmark and our additional synthetic benchmark for modeling a more diverse range of UFOs. The proposed framework consistently enhances performance by a large margin across all four baseline detectors: SECOND, PointPillars, PV-RCNN, and PartA2, giving insight for future work on UFO detection in the wild. △ Less

Submitted 8 January, 2024; originally announced January 2024.

Comments: Under review

Blind-Touch: Homomorphic Encryption-Based Distributed Neural Network Inference for Privacy-Preserving Fingerprint Authentication

Authors: Hyunmin Choi, Simon Woo, Hyoungshick Kim

Abstract: Fingerprint authentication is a popular security mechanism for smartphones and laptops. However, its adoption in web and cloud environments has been limited due to privacy concerns over storing and processing biometric data on servers. This paper introduces Blind-Touch, a novel machine learning-based fingerprint authentication system leveraging homomorphic encryption to address these privacy conce… ▽ More Fingerprint authentication is a popular security mechanism for smartphones and laptops. However, its adoption in web and cloud environments has been limited due to privacy concerns over storing and processing biometric data on servers. This paper introduces Blind-Touch, a novel machine learning-based fingerprint authentication system leveraging homomorphic encryption to address these privacy concerns. Homomorphic encryption allows computations on encrypted data without decrypting. Thus, Blind-Touch can keep fingerprint data encrypted on the server while performing machine learning operations. Blind-Touch combines three strategies to efficiently utilize homomorphic encryption in machine learning: (1) It optimizes the feature vector for a distributed architecture, processing the first fully connected layer (FC-16) in plaintext on the client side and the subsequent layer (FC-1) post-encryption on the server, thereby minimizing encrypted computations; (2) It employs a homomorphic encryption compatible data compression technique capable of handling 8,192 authentication results concurrently; and (3) It utilizes a clustered server architecture to simultaneously process authentication results, thereby enhancing scalability with increasing user numbers. Blind-Touch achieves high accuracy on two benchmark fingerprint datasets, with a 93.6% F1- score for the PolyU dataset and a 98.2% F1-score for the SOKOTO dataset. Moreover, Blind-Touch can match a fingerprint among 5,000 in about 0.65 seconds. With its privacy focused design, high accuracy, and efficiency, Blind-Touch is a promising alternative to conventional fingerprint authentication for web and cloud applications. △ Less

Submitted 1 April, 2024; v1 submitted 18 December, 2023; originally announced December 2023.

Comments: The 38th Annual AAAI Conference on Artificial Intelligence (AAAI) 2024

FineControlNet: Fine-level Text Control for Image Generation with Spatially Aligned Text Control Injection

Authors: Hongsuk Choi, Isaac Kasahara, Selim Engin, Moritz Graule, Nikhil Chavan-Dafle, Volkan Isler

Abstract: Recently introduced ControlNet has the ability to steer the text-driven image generation process with geometric input such as human 2D pose, or edge features. While ControlNet provides control over the geometric form of the instances in the generated image, it lacks the capability to dictate the visual appearance of each instance. We present FineControlNet to provide fine control over each instanc… ▽ More Recently introduced ControlNet has the ability to steer the text-driven image generation process with geometric input such as human 2D pose, or edge features. While ControlNet provides control over the geometric form of the instances in the generated image, it lacks the capability to dictate the visual appearance of each instance. We present FineControlNet to provide fine control over each instance's appearance while maintaining the precise pose control capability. Specifically, we develop and demonstrate FineControlNet with geometric control via human pose images and appearance control via instance-level text prompts. The spatial alignment of instance-specific text prompts and 2D poses in latent space enables the fine control capabilities of FineControlNet. We evaluate the performance of FineControlNet with rigorous comparison against state-of-the-art pose-conditioned text-to-image diffusion models. FineControlNet achieves superior performance in generating images that follow the user-provided instance-specific text prompts and poses compared with existing methods. Project webpage: https://samsunglabs.github.io/FineControlNet-project-page △ Less

Submitted 14 December, 2023; originally announced December 2023.

Comments: Hongsuk Choi and Isaac Kasahara have eqaul contributions. 19 pages, 15 figures, 3 tables

A Unified Multi-Phase CT Synthesis and Classification Framework for Kidney Cancer Diagnosis with Incomplete Data

Authors: Kwang-Hyun Uhm, Seung-Won Jung, Moon Hyung Choi, Sung-Hoo Hong, Sung-Jea Ko

Abstract: Multi-phase CT is widely adopted for the diagnosis of kidney cancer due to the complementary information among phases. However, the complete set of multi-phase CT is often not available in practical clinical applications. In recent years, there have been some studies to generate the missing modality image from the available data. Nevertheless, the generated images are not guaranteed to be effectiv… ▽ More Multi-phase CT is widely adopted for the diagnosis of kidney cancer due to the complementary information among phases. However, the complete set of multi-phase CT is often not available in practical clinical applications. In recent years, there have been some studies to generate the missing modality image from the available data. Nevertheless, the generated images are not guaranteed to be effective for the diagnosis task. In this paper, we propose a unified framework for kidney cancer diagnosis with incomplete multi-phase CT, which simultaneously recovers missing CT images and classifies cancer subtypes using the completed set of images. The advantage of our framework is that it encourages a synthesis model to explicitly learn to generate missing CT phases that are helpful for classifying cancer subtypes. We further incorporate lesion segmentation network into our framework to exploit lesion-level features for effective cancer classification in the whole CT volumes. The proposed framework is based on fully 3D convolutional neural networks to jointly optimize both synthesis and classification of 3D CT volumes. Extensive experiments on both in-house and external datasets demonstrate the effectiveness of our framework for the diagnosis with incomplete data compared with state-of-the-art baselines. In particular, cancer subtype classification using the completed CT data by our method achieves higher performance than the classification using the given incomplete data. △ Less

Submitted 9 December, 2023; originally announced December 2023.

Comments: This article has been accepted for publication in IEEE Journal of Biomedical and Health Informatics

Journal ref: JBHI, 2022

ProsDectNet: Bridging the Gap in Prostate Cancer Detection via Transrectal B-mode Ultrasound Imaging

Authors: Sulaiman Vesal, Indrani Bhattacharya, Hassan Jahanandish, Xinran Li, Zachary Kornberg, Steve Ran Zhou, Elijah Richard Sommer, Moon Hyung Choi, Richard E. Fan, Geoffrey A. Sonn, Mirabela Rusu

Abstract: Interpreting traditional B-mode ultrasound images can be challenging due to image artifacts (e.g., shadowing, speckle), leading to low sensitivity and limited diagnostic accuracy. While Magnetic Resonance Imaging (MRI) has been proposed as a solution, it is expensive and not widely available. Furthermore, most biopsies are guided by Transrectal Ultrasound (TRUS) alone and can miss up to 52% cancer… ▽ More Interpreting traditional B-mode ultrasound images can be challenging due to image artifacts (e.g., shadowing, speckle), leading to low sensitivity and limited diagnostic accuracy. While Magnetic Resonance Imaging (MRI) has been proposed as a solution, it is expensive and not widely available. Furthermore, most biopsies are guided by Transrectal Ultrasound (TRUS) alone and can miss up to 52% cancers, highlighting the need for improved targeting. To address this issue, we propose ProsDectNet, a multi-task deep learning approach that localizes prostate cancer on B-mode ultrasound. Our model is pre-trained using radiologist-labeled data and fine-tuned using biopsy-confirmed labels. ProsDectNet includes a lesion detection and patch classification head, with uncertainty minimization using entropy to improve model performance and reduce false positive predictions. We trained and validated ProsDectNet using a cohort of 289 patients who underwent MRI-TRUS fusion targeted biopsy. We then tested our approach on a group of 41 patients and found that ProsDectNet outperformed the average expert clinician in detecting prostate cancer on B-mode ultrasound images, achieving a patient-level ROC-AUC of 82%, a sensitivity of 74%, and a specificity of 67%. Our results demonstrate that ProsDectNet has the potential to be used as a computer-aided diagnosis system to improve targeted biopsy and treatment planning. △ Less

Submitted 8 December, 2023; originally announced December 2023.

Comments: Accepted in NeurIPS 2023 (Medical Imaging meets NeurIPS Workshop)

Information divergences of Markov chains and their applications

Authors: Youjia Wang, Michael C. H. Choi

Abstract: In this paper, we first introduce and define several new information divergences in the space of transition matrices of finite Markov chains which measure the discrepancy between two Markov chains. These divergences offer natural generalizations of classical information-theoretic divergences, such as the $f$-divergences and the Rényi divergence between probability measures, to the context of finit… ▽ More In this paper, we first introduce and define several new information divergences in the space of transition matrices of finite Markov chains which measure the discrepancy between two Markov chains. These divergences offer natural generalizations of classical information-theoretic divergences, such as the $f$-divergences and the Rényi divergence between probability measures, to the context of finite Markov chains. We begin by detailing and deriving fundamental properties of these divergences and notably gives a Markov chain version of the Pinsker's inequality and Chernoff information. We then utilize these notions in a few applications. First, we investigate the binary hypothesis testing problem of Markov chains, where the newly defined Rényi divergence between Markov chains and its geometric interpretation play an important role in the analysis. Second, we propose and analyze information-theoretic (Cesàro) mixing times and ergodicity coefficients, along with spectral bounds of these notions in the reversible setting. Examples of the random walk on the hypercube, as well as the connections between the critical height of the low-temperature Metropolis-Hastings chain and these proposed ergodicity coefficients, are highlighted. △ Less

Submitted 8 December, 2023; originally announced December 2023.

Comments: 36 pages

MSC Class: 60J10; 60J20; 94A15; 94A17

Explore, Select, Derive, and Recall: Augmenting LLM with Human-like Memory for Mobile Task Automation

Authors: Sunjae Lee, Junyoung Choi, Jungjae Lee, Munim Hasan Wasi, Hojun Choi, Steven Y. Ko, Sangeun Oh, Insik Shin

Abstract: The advent of large language models (LLMs) has opened up new opportunities in the field of mobile task automation. Their superior language understanding and reasoning capabilities allow users to automate complex and repetitive tasks. However, due to the inherent unreliability and high operational cost of LLMs, their practical applicability is quite limited. To address these issues, this paper intr… ▽ More The advent of large language models (LLMs) has opened up new opportunities in the field of mobile task automation. Their superior language understanding and reasoning capabilities allow users to automate complex and repetitive tasks. However, due to the inherent unreliability and high operational cost of LLMs, their practical applicability is quite limited. To address these issues, this paper introduces MobileGPT, an innovative LLM-based mobile task automator equipped with a human-like app memory. MobileGPT emulates the cognitive process of humans interacting with a mobile app -- explore, select, derive, and recall. This approach allows for a more precise and efficient learning of a task's procedure by breaking it down into smaller, modular sub-tasks that can be re-used, re-arranged, and adapted for various objectives. We implement MobileGPT using online LLMs services (GPT-3.5 and GPT-4) and evaluate its performance on a dataset of 160 user instructions across 8 widely used mobile apps. The results indicate that MobileGPT can automate and learn new tasks with 82.5% accuracy, and is able to adapt them to different contexts with near perfect (98.75%) accuracy while reducing both latency and cost by 62.5% and 68.8%, respectively, compared to the GPT-4 powered baseline. △ Less

Submitted 16 March, 2024; v1 submitted 4 December, 2023; originally announced December 2023.

Synchronizing Vision and Language: Bidirectional Token-Masking AutoEncoder for Referring Image Segmentation

Authors: Minhyeok Lee, Dogyoon Lee, Jungho Lee, Suhwan Cho, Heeseung Choi, Ig-Jae Kim, Sangyoun Lee

Abstract: Referring Image Segmentation (RIS) aims to segment target objects expressed in natural language within a scene at the pixel level. Various recent RIS models have achieved state-of-the-art performance by generating contextual tokens to model multimodal features from pretrained encoders and effectively fusing them using transformer-based cross-modal attention. While these methods match language feat… ▽ More Referring Image Segmentation (RIS) aims to segment target objects expressed in natural language within a scene at the pixel level. Various recent RIS models have achieved state-of-the-art performance by generating contextual tokens to model multimodal features from pretrained encoders and effectively fusing them using transformer-based cross-modal attention. While these methods match language features with image features to effectively identify likely target objects, they often struggle to correctly understand contextual information in complex and ambiguous sentences and scenes. To address this issue, we propose a novel bidirectional token-masking autoencoder (BTMAE) inspired by the masked autoencoder (MAE). The proposed model learns the context of image-to-language and language-to-image by reconstructing missing features in both image and language features at the token level. In other words, this approach involves mutually complementing across the features of images and language, with a focus on enabling the network to understand interconnected deep contextual information between the two modalities. This learning method enhances the robustness of RIS performance in complex sentences and scenes. Our BTMAE achieves state-of-the-art performance on three popular datasets, and we demonstrate the effectiveness of the proposed method through various ablation studies. △ Less

Submitted 29 November, 2023; originally announced November 2023.

HierSpeech++: Bridging the Gap between Semantic and Acoustic Representation of Speech by Hierarchical Variational Inference for Zero-shot Speech Synthesis

Authors: Sang-Hoon Lee, Ha-Yeong Choi, Seung-Bin Kim, Seong-Whan Lee

Abstract: Large language models (LLM)-based speech synthesis has been widely adopted in zero-shot speech synthesis. However, they require a large-scale data and possess the same limitations as previous autoregressive speech models, including slow inference speed and lack of robustness. This paper proposes HierSpeech++, a fast and strong zero-shot speech synthesizer for text-to-speech (TTS) and voice convers… ▽ More Large language models (LLM)-based speech synthesis has been widely adopted in zero-shot speech synthesis. However, they require a large-scale data and possess the same limitations as previous autoregressive speech models, including slow inference speed and lack of robustness. This paper proposes HierSpeech++, a fast and strong zero-shot speech synthesizer for text-to-speech (TTS) and voice conversion (VC). We verified that hierarchical speech synthesis frameworks could significantly improve the robustness and expressiveness of the synthetic speech. Furthermore, we significantly improve the naturalness and speaker similarity of synthetic speech even in zero-shot speech synthesis scenarios. For text-to-speech, we adopt the text-to-vec framework, which generates a self-supervised speech representation and an F0 representation based on text representations and prosody prompts. Then, HierSpeech++ generates speech from the generated vector, F0, and voice prompt. We further introduce a high-efficient speech super-resolution framework from 16 kHz to 48 kHz. The experimental results demonstrated that the hierarchical variational autoencoder could be a strong zero-shot speech synthesizer given that it outperforms LLM-based and diffusion-based models. Moreover, we achieved the first human-level quality zero-shot speech synthesis. Audio samples and source code are available at https://github.com/sh-lee-prml/HierSpeechpp. △ Less

Submitted 27 November, 2023; v1 submitted 21 November, 2023; originally announced November 2023.

Comments: 16 pages, 9 figures, 12 tables

Path Planning in 3D with Motion Primitives for Wind Energy-Harvesting Fixed-Wing Aircraft

Authors: Seung-Keol Ryu, Michael Moncton, Han-Lim Choi, Eric Frew

Abstract: In this work, a set of motion primitives is defined for use in an energy-aware motion planning problem. The motion primitives are defined as sequences of control inputs to a simplified four-DOF dynamics model and are used to replace the traditional continuous control space used in many sampling-based motion planners. The primitives are implemented in a Stable Sparse Rapidly Exploring Random Tree (… ▽ More In this work, a set of motion primitives is defined for use in an energy-aware motion planning problem. The motion primitives are defined as sequences of control inputs to a simplified four-DOF dynamics model and are used to replace the traditional continuous control space used in many sampling-based motion planners. The primitives are implemented in a Stable Sparse Rapidly Exploring Random Tree (SST) motion planner and compared to an identical planner using a continuous control space. The planner using primitives was found to run 11.0\% faster but yielded solution paths that were on average worse with higher variance. Also, the solution path travel time is improved by about 50\%. Using motion primitives for sampling spaces in SST can effectively reduce the run time of the algorithm, although at the cost of solution quality. △ Less

Submitted 17 November, 2023; originally announced November 2023.

Comments: 4 pages

Diff-HierVC: Diffusion-based Hierarchical Voice Conversion with Robust Pitch Generation and Masked Prior for Zero-shot Speaker Adaptation

Authors: Ha-Yeong Choi, Sang-Hoon Lee, Seong-Whan Lee

Abstract: Although voice conversion (VC) systems have shown a remarkable ability to transfer voice style, existing methods still have an inaccurate pitch and low speaker adaptation quality. To address these challenges, we introduce Diff-HierVC, a hierarchical VC system based on two diffusion models. We first introduce DiffPitch, which can effectively generate F0 with the target voice style. Subsequently, th… ▽ More Although voice conversion (VC) systems have shown a remarkable ability to transfer voice style, existing methods still have an inaccurate pitch and low speaker adaptation quality. To address these challenges, we introduce Diff-HierVC, a hierarchical VC system based on two diffusion models. We first introduce DiffPitch, which can effectively generate F0 with the target voice style. Subsequently, the generated F0 is fed to DiffVoice to convert the speech with a target voice style. Furthermore, using the source-filter encoder, we disentangle the speech and use the converted Mel-spectrogram as a data-driven prior in DiffVoice to improve the voice style transfer capacity. Finally, by using the masked prior in diffusion models, our model can improve the speaker adaptation quality. Experimental results verify the superiority of our model in pitch generation and voice style transfer performance, and our model also achieves a CER of 0.83% and EER of 3.29% in zero-shot VC scenarios. △ Less

Submitted 8 November, 2023; originally announced November 2023.

Comments: INTERSPEECH 2023 (Oral)