Showing 1–50 of 86 results for author: Hill, M

Life Histories of Taboo Knowledge Artifacts

Authors: Kaylea Champion, Benjamin Mako Hill

Abstract: Communicating about some vital topics -- such as sexuality and health -- is treated as taboo and subjected to censorship. How can we construct knowledge about these topics? Wikipedia is home to numerous high-quality knowledge artifacts about taboo topics like sexual organs and human reproduction. How did these artifacts come into being? How is their existence sustained? This mixed-methods comparat… ▽ More Communicating about some vital topics -- such as sexuality and health -- is treated as taboo and subjected to censorship. How can we construct knowledge about these topics? Wikipedia is home to numerous high-quality knowledge artifacts about taboo topics like sexual organs and human reproduction. How did these artifacts come into being? How is their existence sustained? This mixed-methods comparative project builds on previous work on taboo topics in Wikipedia and draws from qualitative and quantitative approaches. We follow a sequential complementary design, developing a narrative articulation of the life of taboo articles, comparing them to nontaboo articles, and examining some of their quantifiable traits. We find that taboo knowledge artifacts develop through multiple successful collaboration styles and, unsurprisingly, that taboo subjects are the sites of conflict. We identify and describe six themes in the development of taboo knowledge artifacts. These artifacts need <i>resilient leadership</i> and <i>engaged organizations</i> to thrive under conditions of <i>limited identifiability</i> and <i>disjointed sensemaking</i>, while contributors simultaneously engage in <i>emergent governance</i> and <i>imagining public audiences</i>. Our observations have important implications for supporting public knowledge work on controversial subjects such as taboos and more generally. △ Less

Submitted 28 August, 2024; originally announced August 2024.

Towards Open Domain Text-Driven Synthesis of Multi-Person Motions

Authors: Mengyi Shan, Lu Dong, Yutao Han, Yuan Yao, Tao Liu, Ifeoma Nwogu, Guo-Jun Qi, Mitch Hill

Abstract: This work aims to generate natural and diverse group motions of multiple humans from textual descriptions. While single-person text-to-motion generation is extensively studied, it remains challenging to synthesize motions for more than one or two subjects from in-the-wild prompts, mainly due to the lack of available datasets. In this work, we curate human pose and motion datasets by estimating pos… ▽ More This work aims to generate natural and diverse group motions of multiple humans from textual descriptions. While single-person text-to-motion generation is extensively studied, it remains challenging to synthesize motions for more than one or two subjects from in-the-wild prompts, mainly due to the lack of available datasets. In this work, we curate human pose and motion datasets by estimating pose information from large-scale image and video datasets. Our models use a transformer-based diffusion framework that accommodates multiple datasets with any number of subjects or frames. Experiments explore both generation of multi-person static poses and generation of multi-person motion sequences. To our knowledge, our method is the first to generate multi-subject motion sequences with high diversity and fidelity from a large variety of textual prompts. △ Less

Submitted 15 July, 2024; v1 submitted 28 May, 2024; originally announced May 2024.

Comments: ECCV 2024. Project page: https://shanmy.github.io/Multi-Motion/

Challenges in Restructuring Community-based Moderation

Authors: Chau Tran, Kejsi Take, Kaylea Champion, Benjamin Mako Hill, Rachel Greenstadt

Abstract: Content moderation practices and technologies need to change over time as requirements and community expectations shift. However, attempts to restructure existing moderation practices can be difficult, especially for platforms that rely on their communities to conduct moderation activities, because changes can transform the workflow and workload of moderators and contributors' reward systems. Thro… ▽ More Content moderation practices and technologies need to change over time as requirements and community expectations shift. However, attempts to restructure existing moderation practices can be difficult, especially for platforms that rely on their communities to conduct moderation activities, because changes can transform the workflow and workload of moderators and contributors' reward systems. Through the study of extensive archival discussions around a prepublication moderation technology on Wikipedia named Flagged Revisions, complemented by seven semi-structured interviews, we identify various challenges in restructuring community-based moderation practices. We learn that while a new system might sound good in theory and perform well in terms of quantitative metrics, it may conflict with existing social norms. Our findings also highlight how the intricate relationship between platforms and self-governed communities can hinder the ability to assess the performance of any new system and introduce considerable costs related to maintaining, overhauling, or scrapping any piece of infrastructure. △ Less

Submitted 27 February, 2024; originally announced February 2024.

Sources of Underproduction in Open Source Software

Authors: Kaylea Champion, Benjamin Mako Hill

Abstract: Because open source software relies on individuals who select their own tasks, it is often underproduced -- a term used by software engineering researchers to describe when a piece of software's relative quality is lower than its relative importance. We examine the social and technical factors associated with underproduction through a comparison of software packaged by the Debian GNU/Linux communi… ▽ More Because open source software relies on individuals who select their own tasks, it is often underproduced -- a term used by software engineering researchers to describe when a piece of software's relative quality is lower than its relative importance. We examine the social and technical factors associated with underproduction through a comparison of software packaged by the Debian GNU/Linux community. We test a series of hypotheses developed from a reading of prior research in software engineering. Although we find that software age and programming language age offer a partial explanation for variation in underproduction, we were surprised to find that the association between underproduction and package age is weaker at high levels of programming language age. With respect to maintenance efforts, we find that additional resources are not always tied to better outcomes. In particular, having higher numbers of contributors is associated with higher underproduction risk. Also, contrary to our expectations, maintainer turnover and maintenance by a declared team are not associated with lower rates of underproduction. Finally, we find that the people working on bugs in underproduced packages tend to be those who are more central to the community's collaboration network structure, although contributors' betweenness centrality (often associated with brokerage in social networks) is not associated with underproduction. △ Less

Submitted 20 January, 2024; originally announced January 2024.

OmniMotionGPT: Animal Motion Generation with Limited Data

Authors: Zhangsihao Yang, Mingyuan Zhou, Mengyi Shan, Bingbing Wen, Ziwei Xuan, Mitch Hill, Junjie Bai, Guo-Jun Qi, Yalin Wang

Abstract: Our paper aims to generate diverse and realistic animal motion sequences from textual descriptions, without a large-scale animal text-motion dataset. While the task of text-driven human motion synthesis is already extensively studied and benchmarked, it remains challenging to transfer this success to other skeleton structures with limited data. In this work, we design a model architecture that imi… ▽ More Our paper aims to generate diverse and realistic animal motion sequences from textual descriptions, without a large-scale animal text-motion dataset. While the task of text-driven human motion synthesis is already extensively studied and benchmarked, it remains challenging to transfer this success to other skeleton structures with limited data. In this work, we design a model architecture that imitates Generative Pretraining Transformer (GPT), utilizing prior knowledge learned from human data to the animal domain. We jointly train motion autoencoders for both animal and human motions and at the same time optimize through the similarity scores among human motion encoding, animal motion encoding, and text CLIP embedding. Presenting the first solution to this problem, we are able to generate animal motions with high diversity and fidelity, quantitatively and qualitatively outperforming the results of training human motion generation baselines on animal data. Additionally, we introduce AnimalML3D, the first text-animal motion dataset with 1240 animation sequences spanning 36 different animal identities. We hope this dataset would mediate the data scarcity problem in text-driven animal motion generation, providing a new playground for the research community. △ Less

Submitted 30 November, 2023; originally announced November 2023.

Comments: The project page is at https://zshyang.github.io/omgpt-website/

Governance Capture in a Self-Governing Community: A Qualitative Comparison of the Serbo-Croatian Wikipedias

Authors: Zarine Kharazian, Kate Starbird, Benjamin Mako Hill

Abstract: What types of governance arrangements makes some self-governed online groups more vulnerable to disinformation campaigns? To answer this question, we present a qualitative comparative analysis of the Croatian and Serbian Wikipedia editions. We do so because between at least 2011 and 2020, the Croatian language version of Wikipedia was taken over by a small group of administrators who introduced fa… ▽ More What types of governance arrangements makes some self-governed online groups more vulnerable to disinformation campaigns? To answer this question, we present a qualitative comparative analysis of the Croatian and Serbian Wikipedia editions. We do so because between at least 2011 and 2020, the Croatian language version of Wikipedia was taken over by a small group of administrators who introduced far-right bias and outright disinformation; dissenting editorial voices were reverted, banned, and blocked. Although Serbian Wikipedia is roughly similar in size and age, shares many linguistic and cultural features, and faced similar threats, it seems to have largely avoided this fate. Based on a grounded theory analysis of interviews with members of both communities and others in cross-functional platform-level roles, we propose that the convergence of three features -- high perceived value as a target, limited early bureaucratic openness, and a preference for personalistic, informal forms of organization over formal ones -- produced a window of opportunity for governance capture on Croatian Wikipedia. Our findings illustrate that online community governing infrastructures can play a crucial role in systematic disinformation campaigns and other influence operations. △ Less

Submitted 6 November, 2023; originally announced November 2023.

Comments: 26 pages, 2 figures. Accepted for publication in Proceedings of the ACM on Human-Computer Interaction (CSCW 2024)

Journal ref: CSCW 2024, Volume 8, Article 61, 1-26

Open Problems in DAOs

Authors: Joshua Tan, Tara Merk, Sarah Hubbard, Eliza R. Oak, Helena Rong, Joni Pirovich, Ellie Rennie, Rolf Hoefer, Michael Zargham, Jason Potts, Chris Berg, Reuben Youngblom, Primavera De Filippi, Seth Frey, Jeff Strnad, Morshed Mannan, Kelsie Nabben, Silke Noa Elrifai, Jake Hartnell, Benjamin Mako Hill, Tobin South, Ryan L. Thomas, Jonathan Dotan, Ariana Spring, Alexia Maddox , et al. (4 additional authors not shown)

Abstract: Decentralized autonomous organizations (DAOs) are a new, rapidly-growing class of organizations governed by smart contracts. Here we describe how researchers can contribute to the emerging science of DAOs and other digitally-constituted organizations. From granular privacy primitives to mechanism designs to model laws, we identify high-impact problems in the DAO ecosystem where existing gaps might… ▽ More Decentralized autonomous organizations (DAOs) are a new, rapidly-growing class of organizations governed by smart contracts. Here we describe how researchers can contribute to the emerging science of DAOs and other digitally-constituted organizations. From granular privacy primitives to mechanism designs to model laws, we identify high-impact problems in the DAO ecosystem where existing gaps might be tackled through a new data set or by applying tools and ideas from existing research fields such as political science, computer science, economics, law, and organizational science. Our recommendations encompass exciting research questions as well as promising business opportunities. We call on the wider research community to join the global effort to invent the next generation of organizations. △ Less

Submitted 12 June, 2024; v1 submitted 29 October, 2023; originally announced October 2023.

Comments: includes major coordination problems

Taboo and Collaborative Knowledge Production: Evidence from Wikipedia

Authors: Kaylea Champion, Benjamin Mako Hill

Abstract: By definition, people are reticent or even unwilling to talk about taboo subjects. Because subjects like sexuality, health, and violence are taboo in most cultures, important information on each of these subjects can be difficult to obtain. Are peer produced knowledge bases like Wikipedia a promising approach for providing people with information on taboo subjects? With its reliance on volunteers… ▽ More By definition, people are reticent or even unwilling to talk about taboo subjects. Because subjects like sexuality, health, and violence are taboo in most cultures, important information on each of these subjects can be difficult to obtain. Are peer produced knowledge bases like Wikipedia a promising approach for providing people with information on taboo subjects? With its reliance on volunteers who might also be averse to taboo, can the peer production model produce high-quality information on taboo subjects? In this paper, we seek to understand the role of taboo in knowledge bases produced by volunteers. We do so by developing a novel computational approach to identify taboo subjects and by using this method to identify a set of articles on taboo subjects in English Wikipedia. We find that articles on taboo subjects are more popular than non-taboo articles and that they are frequently vandalized. Despite frequent vandalism attacks, we also find that taboo articles are higher quality than non-taboo articles. We hypothesize that stigmatizing societal attitudes will lead contributors to taboo subjects to seek to be less identifiable. Although our results are consistent with this proposal in several ways, we surprisingly find that contributors make themselves more identifiable in others. △ Less

Submitted 11 August, 2023; originally announced August 2023.

Recognizing People by Body Shape Using Deep Networks of Images and Words

Authors: Blake A. Myers, Lucas Jaggernauth, Thomas M. Metz, Matthew Q. Hill, Veda Nandan Gandi, Carlos D. Castillo, Alice J. O'Toole

Abstract: Common and important applications of person identification occur at distances and viewpoints in which the face is not visible or is not sufficiently resolved to be useful. We examine body shape as a biometric across distance and viewpoint variation. We propose an approach that combines standard object classification networks with representations based on linguistic (word-based) descriptions of bod… ▽ More Common and important applications of person identification occur at distances and viewpoints in which the face is not visible or is not sufficiently resolved to be useful. We examine body shape as a biometric across distance and viewpoint variation. We propose an approach that combines standard object classification networks with representations based on linguistic (word-based) descriptions of bodies. Algorithms with and without linguistic training were compared on their ability to identify people from body shape in images captured across a large range of distances/views (close-range, 100m, 200m, 270m, 300m, 370m, 400m, 490m, 500m, 600m, and at elevated pitch in images taken by an unmanned aerial vehicle [UAV]). Accuracy, as measured by identity-match ranking and false accept errors in an open-set test, was surprisingly good. For identity-ranking, linguistic models were more accurate for close-range images, whereas non-linguistic models fared better at intermediary distances. Fusion of the linguistic and non-linguistic embeddings improved performance at all, but the farthest distance. Although the non-linguistic model yielded fewer false accepts at all distances, fusion of the linguistic and non-linguistic models decreased false accepts for all, but the UAV images. We conclude that linguistic and non-linguistic representations of body shape can offer complementary identity information for bodies that can improve identification in applications of interest. △ Less

Submitted 30 May, 2023; originally announced May 2023.

Comments: 9 pages, 5 figures, 4 tables

Neural Level Set Topology Optimization Using Unfitted Finite Elements

Authors: Connor N. Mallon, Aaron W. Thornton, Matthew R. Hill, Santiago Badia

Abstract: To facilitate widespread adoption of automated engineering design techniques, existing methods must become more efficient and generalizable. In the field of topology optimization, this requires the coupling of modern optimization methods with solvers capable of handling arbitrary problems. In this work, a topology optimization method for general multiphysics problems is presented. We leverage a co… ▽ More To facilitate widespread adoption of automated engineering design techniques, existing methods must become more efficient and generalizable. In the field of topology optimization, this requires the coupling of modern optimization methods with solvers capable of handling arbitrary problems. In this work, a topology optimization method for general multiphysics problems is presented. We leverage a convolutional neural parameterization of a level set for a description of the geometry and use this in an unfitted finite element method that is differentiable with respect to the level set everywhere in the domain. We construct the parameter to objective map in such a way that the gradient can be computed entirely by automatic differentiation at roughly the cost of an objective function evaluation. The method produces optimized topologies that are similar in performance yet exhibit greater regularity than baseline approaches on standard benchmarks whilst having the ability to solve a more general class of problems, e.g., interface-coupled multiphysics. △ Less

Submitted 22 February, 2024; v1 submitted 23 March, 2023; originally announced March 2023.

Comments: 16 pages + refs, 10 figs

Coding Guidelines and Undecidability

Authors: Roberto Bagnara, Abramo Bagnara, Patricia M. Hill

Abstract: The C and C++ programming languages are widely used for the implementation of software in critical systems. They are complex languages with subtle features and peculiarities that might baffle even the more expert programmers. Hence, the general prescription of language subsetting, which occurs in most functional safety standards and amounts to only using a "safer" subset of the language, is partic… ▽ More The C and C++ programming languages are widely used for the implementation of software in critical systems. They are complex languages with subtle features and peculiarities that might baffle even the more expert programmers. Hence, the general prescription of language subsetting, which occurs in most functional safety standards and amounts to only using a "safer" subset of the language, is particularly applicable to them. Coding guidelines are the preferred way of expressing language subsets. Some guidelines are formulated in terms of the programming language and its implementation only: in this case they are amenable to automatic checking. However, due to fundamental limitations of computing, some guidelines are undecidable, that is, they are based on program properties that no current and future algorithm can capture in all cases. The most mature and widespread coding standards, the MISRA ones, explicitly tag guidelines with undecidable or decidable. It turns out that this information is not of secondary nature and must be taken into account for a full understanding of what the guideline is asking for. As a matter of fact, undecidability is a common source of confusion affecting many users of coding standards and of the associated checking tools. In this paper, we recall the notions of decidability and undecidability in terms that are understandable to any C/C++ programmer. The paper includes a systematic study of all the undecidable MISRA C:2012 guidelines, discussing the reasons for the undecidability and its consequences. We pay particular attention to undecidable guidelines that have decidable approximations whose enforcement would not overly constrain the source code. We also discuss some coding guidelines for which compliance is hard, if not impossible, to prove, even beyond the issue of decidability. △ Less

Submitted 28 December, 2022; originally announced December 2022.

Comments: 12 pages, 5 figures, 1 table

MSC Class: 68N15 ACM Class: F.3.1; D.3.0

Many Destinations, Many Pathways: A Quantitative Analysis of Legitimate Peripheral Participation in Scratch

Authors: Ruijia Cheng, Benjamin Mako Hill

Abstract: Although informal online learning communities have proliferated over the last two decades, a fundamental question remains: What are the users of these communities expected to learn? Guided by the work of Etienne Wenger on communities of practice, we identify three distinct types of learning goals common to online informal learning communities: the development of domain skills, the development of i… ▽ More Although informal online learning communities have proliferated over the last two decades, a fundamental question remains: What are the users of these communities expected to learn? Guided by the work of Etienne Wenger on communities of practice, we identify three distinct types of learning goals common to online informal learning communities: the development of domain skills, the development of identity as a community member, and the development of community-specific values and practices. Given these goals, what is the best way to support learning? Drawing from previous research in social computing, we ask how different types of legitimate peripheral participation by newcomers-contribution to core tasks, engagement with practice proxies, social bonding, and feedback exchange-may be associated with these three learning goals. Using data from the Scratch online community, we conduct a quantitative analysis to explore these questions. Our study contributes both theoretical insights and empirical evidence on how different types of learning occur in informal online environments. △ Less

Submitted 8 November, 2022; originally announced November 2022.

Journal ref: Proc. ACM Hum.-Comput. Interact. 6, CSCW2, Article 381 (November 2022), 26 pages

Learning Probabilistic Models from Generator Latent Spaces with Hat EBM

Authors: Mitch Hill, Erik Nijkamp, Jonathan Mitchell, Bo Pang, Song-Chun Zhu

Abstract: This work proposes a method for using any generator network as the foundation of an Energy-Based Model (EBM). Our formulation posits that observed images are the sum of unobserved latent variables passed through the generator network and a residual random variable that spans the gap between the generator output and the image manifold. One can then define an EBM that includes the generator as part… ▽ More This work proposes a method for using any generator network as the foundation of an Energy-Based Model (EBM). Our formulation posits that observed images are the sum of unobserved latent variables passed through the generator network and a residual random variable that spans the gap between the generator output and the image manifold. One can then define an EBM that includes the generator as part of its forward pass, which we call the Hat EBM. The model can be trained without inferring the latent variables of the observed data or calculating the generator Jacobian determinant. This enables explicit probabilistic modeling of the output distribution of any type of generator network. Experiments show strong performance of the proposed method on (1) unconditional ImageNet synthesis at 128x128 resolution, (2) refining the output of existing generators, and (3) learning EBMs that incorporate non-probabilistic generators. Code and pretrained models to reproduce our results are available at https://github.com/point0bar1/hat-ebm. △ Less

Submitted 12 January, 2023; v1 submitted 28 October, 2022; originally announced October 2022.

Comments: NeurIPS 2022 camera ready

EBM Life Cycle: MCMC Strategies for Synthesis, Defense, and Density Modeling

Authors: Mitch Hill, Jonathan Mitchell, Chu Chen, Yuan Du, Mubarak Shah, Song-Chun Zhu

Abstract: This work presents strategies to learn an Energy-Based Model (EBM) according to the desired length of its MCMC sampling trajectories. MCMC trajectories of different lengths correspond to models with different purposes. Our experiments cover three different trajectory magnitudes and learning outcomes: 1) shortrun sampling for image generation; 2) midrun sampling for classifier-agnostic adversarial… ▽ More This work presents strategies to learn an Energy-Based Model (EBM) according to the desired length of its MCMC sampling trajectories. MCMC trajectories of different lengths correspond to models with different purposes. Our experiments cover three different trajectory magnitudes and learning outcomes: 1) shortrun sampling for image generation; 2) midrun sampling for classifier-agnostic adversarial defense; and 3) longrun sampling for principled modeling of image probability densities. To achieve these outcomes, we introduce three novel methods of MCMC initialization for negative samples used in Maximum Likelihood (ML) learning. With standard network architectures and an unaltered ML objective, our MCMC initialization methods alone enable significant performance gains across the three applications that we investigate. Our results include state-of-the-art FID scores for unnormalized image densities on the CIFAR-10 and ImageNet datasets; state-of-the-art adversarial defense on CIFAR-10 among purification methods and the first EBM defense on ImageNet; and scalable techniques for learning valid probability densities. Code for this project can be found at https://github.com/point0bar1/ebm-life-cycle. △ Less

Submitted 24 May, 2022; originally announced May 2022.

Federated Learning Enables Big Data for Rare Cancer Boundary Detection

Authors: Sarthak Pati, Ujjwal Baid, Brandon Edwards, Micah Sheller, Shih-Han Wang, G Anthony Reina, Patrick Foley, Alexey Gruzdev, Deepthi Karkada, Christos Davatzikos, Chiharu Sako, Satyam Ghodasara, Michel Bilello, Suyash Mohan, Philipp Vollmuth, Gianluca Brugnara, Chandrakanth J Preetha, Felix Sahm, Klaus Maier-Hein, Maximilian Zenk, Martin Bendszus, Wolfgang Wick, Evan Calabrese, Jeffrey Rudie, Javier Villanueva-Meyer , et al. (254 additional authors not shown)

Abstract: Although machine learning (ML) has shown promise in numerous domains, there are concerns about generalizability to out-of-sample data. This is currently addressed by centrally sharing ample, and importantly diverse, data from multiple sites. However, such centralization is challenging to scale (or even not feasible) due to various limitations. Federated ML (FL) provides an alternative to train acc… ▽ More Although machine learning (ML) has shown promise in numerous domains, there are concerns about generalizability to out-of-sample data. This is currently addressed by centrally sharing ample, and importantly diverse, data from multiple sites. However, such centralization is challenging to scale (or even not feasible) due to various limitations. Federated ML (FL) provides an alternative to train accurate and generalizable ML models, by only sharing numerical model updates. Here we present findings from the largest FL study to-date, involving data from 71 healthcare institutions across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, utilizing the largest dataset of such patients ever used in the literature (25,256 MRI scans from 6,314 patients). We demonstrate a 33% improvement over a publicly trained model to delineate the surgically targetable tumor, and 23% improvement over the tumor's entire extent. We anticipate our study to: 1) enable more studies in healthcare informed by large and diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further quantitative analyses for glioblastoma via performance optimization of our consensus model for eventual public release, and 3) demonstrate the effectiveness of FL at such scale and task complexity as a paradigm shift for multi-site collaborations, alleviating the need for data sharing. △ Less

Submitted 25 April, 2022; v1 submitted 22 April, 2022; originally announced April 2022.

Comments: federated learning, deep learning, convolutional neural network, segmentation, brain tumor, glioma, glioblastoma, FeTS, BraTS

How Interest-Driven Content Creation Shapes Opportunities for Informal Learning in Scratch: A Case Study on Novices' Use of Data Structures

Authors: Ruijia Cheng, Sayamindu Dasgupta, Benjamin Mako Hill

Abstract: Through a mixed-method analysis of data from Scratch, we examine how novices learn to program with simple data structures by using community-produced learning resources. First, we present a qualitative study that describes how community-produced learning resources create archetypes that shape exploration and may disadvantage some with less common interests. In a second quantitative study, we find… ▽ More Through a mixed-method analysis of data from Scratch, we examine how novices learn to program with simple data structures by using community-produced learning resources. First, we present a qualitative study that describes how community-produced learning resources create archetypes that shape exploration and may disadvantage some with less common interests. In a second quantitative study, we find broad support for this dynamic in several hypothesis tests. Our findings identify a social feedback loop that we argue could limit sources of inspiration, pose barriers to broadening participation, and confine learners' understanding of general concepts. We conclude by suggesting several approaches that may mitigate these dynamics. △ Less

Submitted 22 March, 2022; originally announced March 2022.

Comments: Ruijia Cheng, Sayamindu Dasgupta, and Benjamin Mako Hill. 2022. How Interest-Driven Content Creation Shapes Opportunities for Informal Learning in Scratch: A Case Study on Novices' Use of Data Structures. In CHI Conference on Human Factors in Computing Systems (CHI '22), April 29-May 5, 2022, New Orleans, LA, USA. ACM, New York, NY, USA, 16 pages

Pond: CXL-Based Memory Pooling Systems for Cloud Platforms

Authors: Huaicheng Li, Daniel S. Berger, Stanko Novakovic, Lisa Hsu, Dan Ernst, Pantea Zardoshti, Monish Shah, Samir Rajadnya, Scott Lee, Ishwar Agarwal, Mark D. Hill, Marcus Fontoura, Ricardo Bianchini

Abstract: Public cloud providers seek to meet stringent performance requirements and low hardware cost. A key driver of performance and cost is main memory. Memory pooling promises to improve DRAM utilization and thereby reduce costs. However, pooling is challenging under cloud performance requirements. This paper proposes Pond, the first memory pooling system that both meets cloud performance goals and sig… ▽ More Public cloud providers seek to meet stringent performance requirements and low hardware cost. A key driver of performance and cost is main memory. Memory pooling promises to improve DRAM utilization and thereby reduce costs. However, pooling is challenging under cloud performance requirements. This paper proposes Pond, the first memory pooling system that both meets cloud performance goals and significantly reduces DRAM cost. Pond builds on the Compute Express Link (CXL) standard for load/store access to pool memory and two key insights. First, our analysis of cloud production traces shows that pooling across 8-16 sockets is enough to achieve most of the benefits. This enables a small-pool design with low access latency. Second, it is possible to create machine learning models that can accurately predict how much local and pool memory to allocate to a virtual machine (VM) to resemble same-NUMA-node memory performance. Our evaluation with 158 workloads shows that Pond reduces DRAM costs by 7% with performance within 1-5% of same-NUMA-node VM allocations. △ Less

Submitted 21 October, 2022; v1 submitted 1 March, 2022; originally announced March 2022.

Comments: Update affiliations

The Risks, Benefits, and Consequences of Prepublication Moderation: Evidence from 17 Wikipedia Language Editions

Authors: Chau Tran, Kaylea Champion, Benjamin Mako Hill, Rachel Greenstadt

Abstract: Many online communities rely on postpublication moderation where contributors, even those that are perceived as being risky, are allowed to publish material immediately and where moderation takes place after the fact. An alternative arrangement involves moderating content before publication. A range of communities have argued against prepublication moderation by suggesting that it makes contributi… ▽ More Many online communities rely on postpublication moderation where contributors, even those that are perceived as being risky, are allowed to publish material immediately and where moderation takes place after the fact. An alternative arrangement involves moderating content before publication. A range of communities have argued against prepublication moderation by suggesting that it makes contributing less enjoyable for new members and that it will distract established community members with extra moderation work. We present an empirical analysis of the effects of a prepublication moderation system called FlaggedRevs that was deployed by several Wikipedia language editions. We used panel data from 17 large Wikipedia editions to test a series of hypotheses related to the effect of the system on activity levels and contribution quality. We found that the system was very effective at keeping low-quality contributions from ever becoming visible. Although there is some evidence that the system discouraged participation among users without accounts, our analysis suggests that the system's effects on contribution volume and quality were moderate at most. Our findings imply that concerns regarding the major negative effects of prepublication moderation systems on contribution quality and project productivity may be overstated. △ Less

Submitted 26 August, 2022; v1 submitted 11 February, 2022; originally announced February 2022.

Comments: This paper was submitted to CSCW2 (November 2022)

No Community Can Do Everything: Why People Participate in Similar Online Communities

Authors: Nathan TeBlunthuis, Charles Kiene, Isabella Brown, Laura Alia Levi, Nicole McGinnis, Benjamin Mako Hill

Abstract: Large-scale quantitative analyses have shown that individuals frequently talk to each other about similar things in different online spaces. Why do these overlapping communities exist? We provide an answer grounded in the analysis of 20 interviews with active participants in clusters of highly related subreddits. Within a broad topical area, there are a diversity of benefits an online community ca… ▽ More Large-scale quantitative analyses have shown that individuals frequently talk to each other about similar things in different online spaces. Why do these overlapping communities exist? We provide an answer grounded in the analysis of 20 interviews with active participants in clusters of highly related subreddits. Within a broad topical area, there are a diversity of benefits an online community can confer. These include (a) specific information and discussion, (b) socialization with similar others, and (c) attention from the largest possible audience. A single community cannot meet all three needs. Our findings suggest that topical areas within an online community platform tend to become populated by groups of specialized communities with diverse sizes, topical boundaries, and rules. Compared with any single community, such systems of overlapping communities are able to provide a greater range of benefits. △ Less

Submitted 10 February, 2022; v1 submitted 11 January, 2022; originally announced January 2022.

Comments: Accepted to CSCW 2022

ACM Class: K.4.0

A Rationale-Based Classification of MISRA C Guidelines

Authors: Roberto Bagnara, Abramo Bagnara, Patricia M. Hill

Abstract: MISRA C is the most authoritative language subset for the C programming language that is a de facto standard in several industry sectors where safety and security are of paramount importance. While MISRA C is currently encoded in 175 guidelines (coding rules and directives), it does not coincide with them: proper adoption of MISRA C requires embracing its preventive approach (as opposed to the "bu… ▽ More MISRA C is the most authoritative language subset for the C programming language that is a de facto standard in several industry sectors where safety and security are of paramount importance. While MISRA C is currently encoded in 175 guidelines (coding rules and directives), it does not coincide with them: proper adoption of MISRA C requires embracing its preventive approach (as opposed to the "bug finding" approach) and a documented development process where justifiable non-compliances are authorized and recorded as deviations. MISRA C guidelines are classified along several axes in the official MISRA documents. In this paper, we add to these an orthogonal classification that associates guidelines with their main rationale. The advantages of this new classification are illustrated for different kinds of projects, including those not (yet) having MISRA compliance among their objectives. △ Less

Submitted 23 December, 2021; originally announced December 2021.

Comments: 12 pages, 2 figures

MSC Class: 68N15 ACM Class: F.3.1; D.3.0

The Hidden Costs of Requiring Accounts: Quasi-Experimental Evidence From Peer Production

Authors: Benjamin Mako Hill, Aaron Shaw

Abstract: Online communities, like Wikipedia, produce valuable public information goods. Whereas some of these communities require would-be contributors to create accounts, many do not. Does this requirement catalyze cooperation or inhibit participation? Prior research provides divergent predictions but little causal evidence. We conduct an empirical test using longitudinal data from 136 natural experiments… ▽ More Online communities, like Wikipedia, produce valuable public information goods. Whereas some of these communities require would-be contributors to create accounts, many do not. Does this requirement catalyze cooperation or inhibit participation? Prior research provides divergent predictions but little causal evidence. We conduct an empirical test using longitudinal data from 136 natural experiments where would-be contributors to wikis were suddenly required to log in to contribute. Requiring accounts leads to a small increase in account creation, but reduces both high- and low-quality contributions from registered and unregistered participants. Although the change deters a large portion of low-quality participation, the vast majority of deterred contributions are of higher quality. We conclude that requiring accounts introduces an undertheorized tradeoff for public goods production in interactive communication systems. △ Less

Submitted 20 November, 2021; originally announced November 2021.

Journal ref: Communication Research 48(6): 771-95, 2021

Qualities of Quality: A Tertiary Review of Software Quality Measurement Research

Authors: Kaylea Champion, Sejal Khatri, Benjamin Mako Hill

Abstract: This paper presents a tertiary review of software quality measurement research. To conduct this review, we examined an initial dataset of 7,811 articles and found 75 relevant and high-quality secondary analyses of software quality research. Synthesizing this body of work, we offer an overview of perspectives, measurement approaches, and trends. We identify five distinct perspectives that conceptua… ▽ More This paper presents a tertiary review of software quality measurement research. To conduct this review, we examined an initial dataset of 7,811 articles and found 75 relevant and high-quality secondary analyses of software quality research. Synthesizing this body of work, we offer an overview of perspectives, measurement approaches, and trends. We identify five distinct perspectives that conceptualize quality as heuristic, as maintainability, as a holistic concept, as structural features of software, and as dependability. We also identify three key challenges. First, we find widespread evidence of validity questions with common measures. Second, we observe the application of machine learning methods without adequate evaluation. Third, we observe the use of aging datasets. Finally, from these observations, we sketch a path toward a theoretical framework that will allow software engineering researchers to systematically confront these weaknesses while remaining grounded in the experiences of developers and the real world in which code is ultimately deployed. △ Less

Submitted 28 July, 2021; originally announced July 2021.

Identifying Competition and Mutualism Between Online Groups

Authors: Nathan TeBlunthuis, Benjamin Mako Hill

Abstract: Platforms often host multiple online groups with overlapping topics and members. How can researchers and designers understand how related groups affect each other? Inspired by population ecology, prior research in social computing and human-computer interaction has studied related groups by correlating group size with degrees of overlap in content and membership, but has produced puzzling results:… ▽ More Platforms often host multiple online groups with overlapping topics and members. How can researchers and designers understand how related groups affect each other? Inspired by population ecology, prior research in social computing and human-computer interaction has studied related groups by correlating group size with degrees of overlap in content and membership, but has produced puzzling results: overlap is associated with competition in some contexts but with mutualism in others. We suggest that this inconsistency results from aggregating intergroup relationships into an overall environmental effect that obscures the diversity of competition and mutualism among related groups. Drawing on the framework of community ecology, we introduce a time-series method for inferring competition and mutualism. We then use this framework to inform a large-scale analysis of clusters of subreddits that all have high user overlap. We find that mutualism is more common than competition. △ Less

Submitted 18 January, 2022; v1 submitted 14 July, 2021; originally announced July 2021.

Comments: 10 pages, 6 figures

A National Discovery Cloud: Preparing the US for Global Competitiveness in the New Era of 21st Century Digital Transformation

Authors: Ian Foster, Daniel Lopresti, Bill Gropp, Mark D. Hill, Katie Schuman

Abstract: The nature of computation and its role in our lives have been transformed in the past two decades by three remarkable developments: the emergence of public cloud utilities as a new computing platform; the ability to extract information from enormous quantities of data via machine learning; and the emergence of computational simulation as a research method on par with experimental science. Each dev… ▽ More The nature of computation and its role in our lives have been transformed in the past two decades by three remarkable developments: the emergence of public cloud utilities as a new computing platform; the ability to extract information from enormous quantities of data via machine learning; and the emergence of computational simulation as a research method on par with experimental science. Each development has major implications for how societies function and compete; together, they represent a change in technological foundations of society as profound as the telegraph or electrification. Societies that embrace these changes will lead in the 21st Century; those that do not, will decline in prosperity and influence. Nowhere is this stark choice more evident than in research and education, the two sectors that produce the innovations that power the future and prepare a workforce able to exploit those innovations, respectively. In this article, we introduce these developments and suggest steps that the US government might take to prepare the research and education system for its implications. △ Less

Submitted 19 April, 2021; v1 submitted 14 April, 2021; originally announced April 2021.

Comments: A Computing Community Consortium (CCC) white paper, 6 pages

Report number: ccc2021whitepaper_4

Underproduction: An Approach for Measuring Risk in Open Source Software

Authors: Kaylea Champion, Benjamin Mako Hill

Abstract: The widespread adoption of Free/Libre and Open Source Software (FLOSS) means that the ongoing maintenance of many widely used software components relies on the collaborative effort of volunteers who set their own priorities and choose their own tasks. We argue that this has created a new form of risk that we call 'underproduction' which occurs when the supply of software engineering labor becomes… ▽ More The widespread adoption of Free/Libre and Open Source Software (FLOSS) means that the ongoing maintenance of many widely used software components relies on the collaborative effort of volunteers who set their own priorities and choose their own tasks. We argue that this has created a new form of risk that we call 'underproduction' which occurs when the supply of software engineering labor becomes out of alignment with the demand of people who rely on the software produced. We present a conceptual framework for identifying relative underproduction in software as well as a statistical method for applying our framework to a comprehensive dataset from the Debian GNU/Linux distribution that includes 21,902 source packages and the full history of 461,656 bugs. We draw on this application to present two experiments: (1) a demonstration of how our technique can be used to identify at-risk software packages in a large FLOSS repository and (2) a validation of these results using an alternate indicator of package risk. Our analysis demonstrates both the utility of our approach and reveals the existence of widespread underproduction in a range of widely-installed software components in Debian. △ Less

Submitted 27 February, 2021; originally announced March 2021.

Comments: Preprint of archival paper accepted for SANER 2021

Accelerating amorphous polymer electrolyte screening by learning to reduce errors in molecular dynamics simulated properties

Authors: Tian Xie, Arthur France-Lanord, Yanming Wang, Jeffrey Lopez, Michael Austin Stolberg, Megan Hill, Graham Michael Leverick, Rafael Gomez-Bombarelli, Jeremiah A. Johnson, Yang Shao-Horn, Jeffrey C. Grossman

Abstract: Polymer electrolytes are promising candidates for the next generation lithium-ion battery technology. Large scale screening of polymer electrolytes is hindered by the significant cost of molecular dynamics (MD) simulation in amorphous systems: the amorphous structure of polymers requires multiple, repeated sampling to reduce noise and the slow relaxation requires long simulation time for convergen… ▽ More Polymer electrolytes are promising candidates for the next generation lithium-ion battery technology. Large scale screening of polymer electrolytes is hindered by the significant cost of molecular dynamics (MD) simulation in amorphous systems: the amorphous structure of polymers requires multiple, repeated sampling to reduce noise and the slow relaxation requires long simulation time for convergence. Here, we accelerate the screening with a multi-task graph neural network that learns from a large amount of noisy, unconverged, short MD data and a small number of converged, long MD data. We achieve accurate predictions of 4 different converged properties and screen a space of 6247 polymers that is orders of magnitude larger than previous computational studies. Further, we extract several design principles for polymer electrolytes and provide an open dataset for the community. Our approach could be applicable to a broad class of material discovery problems that involve the simulation of complex, amorphous materials. △ Less

Submitted 15 March, 2022; v1 submitted 13 January, 2021; originally announced January 2021.

Comments: 29 pages, 6 figures + supplementary information

Journal ref: Nature communications 13.1 (2022): 1-10

Advancing Computing's Foundation of US Industry & Society

Authors: Thomas M. Conte, Ian T. Foster, William Gropp, Mark D. Hill

Abstract: While past information technology (IT) advances have transformed society, future advances hold even greater promise. For example, we have only just begun to reap the changes from artificial intelligence (AI), especially machine learning (ML). Underlying IT's impact are the dramatic improvements in computer hardware, which deliver performance that unlock new capabilities. For example, recent succes… ▽ More While past information technology (IT) advances have transformed society, future advances hold even greater promise. For example, we have only just begun to reap the changes from artificial intelligence (AI), especially machine learning (ML). Underlying IT's impact are the dramatic improvements in computer hardware, which deliver performance that unlock new capabilities. For example, recent successes in AI/ML required the synergy of improved algorithms and hardware architectures (e.g., general-purpose graphics processing units). However, unlike in the 20th Century and early 2000s, tomorrow's performance aspirations must be achieved without continued semiconductor scaling formerly provided by Moore's Law and Dennard Scaling. How will one deliver the next 100x improvement in capability at similar or less cost to enable great value? Can we make the next AI leap without 100x better hardware? This whitepaper argues for a multipronged effort to develop new computing approaches beyond Moore's Law to advance the foundation that computing provides to US industry, education, medicine, science, and government. This impact extends far beyond the IT industry itself, as IT is now central for providing value across society, for example in semi-autonomous vehicles, tele-education, health wearables, viral analysis, and efficient administration. Herein we draw upon considerable visioning work by CRA's Computing Community Consortium (CCC) and the IEEE Rebooting Computing Initiative (IEEE RCI), enabled by thought leader input from industry, academia, and the US government. △ Less

Submitted 4 January, 2021; originally announced January 2021.

Comments: A Computing Community Consortium (CCC) white paper, 4 pages

Report number: ccc2020whitepaper_17

Designing for Critical Algorithmic Literacies

Authors: Sayamindu Dasgupta, Benjamin Mako Hill

Abstract: As pervasive data collection and powerful algorithms increasingly shape children's experience of the world and each other, their ability to interrogate computational algorithms has become crucially important. A growing body of work has attempted to articulate a set of "literacies" to describe the intellectual tools that children can use to understand, interrogate, and critique the algorithmic syst… ▽ More As pervasive data collection and powerful algorithms increasingly shape children's experience of the world and each other, their ability to interrogate computational algorithms has become crucially important. A growing body of work has attempted to articulate a set of "literacies" to describe the intellectual tools that children can use to understand, interrogate, and critique the algorithmic systems that shape their lives. Unfortunately, because many algorithms are invisible, only a small number of children develop the literacies required to critique these systems. How might designers support the development of critical algorithmic literacies? Based on our experience designing two data programming systems, we present four design principles that we argue can help children develop literacies that allow them to understand not only how algorithms work, but also to critique and question them. △ Less

Submitted 4 August, 2020; originally announced August 2020.

ACM Class: H.1.2; K.3.1; K.3.2

Opportunities and Challenges for Next Generation Computing

Authors: Gregory D. Hager, Mark D. Hill, Katherine Yelick

Abstract: Computing has dramatically changed nearly every aspect of our lives, from business and agriculture to communication and entertainment. As a nation, we rely on computing in the design of systems for energy, transportation and defense; and computing fuels scientific discoveries that will improve our fundamental understanding of the world and help develop solutions to major challenges in health and t… ▽ More Computing has dramatically changed nearly every aspect of our lives, from business and agriculture to communication and entertainment. As a nation, we rely on computing in the design of systems for energy, transportation and defense; and computing fuels scientific discoveries that will improve our fundamental understanding of the world and help develop solutions to major challenges in health and the environment. Computing has changed our world, in part, because our innovations can run on computers whose performance and cost-performance has improved a million-fold over the last few decades. A driving force behind this has been a repeated doubling of the transistors per chip, dubbed Moore's Law. A concomitant enabler has been Dennard Scaling that has permitted these performance doublings at roughly constant power, but, as we will see, both trends face challenges. Consider for a moment the impact of these two trends over the past 30 years. A 1980's supercomputer (e.g. a Cray 2) was rated at nearly 2 Gflops and consumed nearly 200 KW of power. At the time, it was used for high performance and national-scale applications ranging from weather forecasting to nuclear weapons research. A computer of similar performance now fits in our pocket and consumes less than 10 watts. What would be the implications of a similar computing/power reduction over the next 30 years - that is, taking a petaflop-scale machine (e.g. the Cray XK7 which requires about 500 KW for 1 Pflop (=1015 operations/sec) performance) and repeating that process? What is possible with such a computer in your pocket? How would it change the landscape of high capacity computing? In the remainder of this paper, we articulate some opportunities and challenges for dramatic performance improvements of both personal to national scale computing, and discuss some "out of the box" possibilities for achieving computing at this scale. △ Less

Submitted 31 July, 2020; originally announced August 2020.

Comments: A Computing Community Consortium (CCC) white paper, 7 pages

Effects of algorithmic flagging on fairness: quasi-experimental evidence from Wikipedia

Authors: Nathan TeBlunthuis, Benjamin Mako Hill, Aaron Halfaker

Abstract: Online community moderators often rely on social signals such as whether or not a user has an account or a profile page as clues that users may cause problems. Reliance on these clues can lead to overprofiling bias when moderators focus on these signals but overlook the misbehavior of others. We propose that algorithmic flagging systems deployed to improve the efficiency of moderation work can als… ▽ More Online community moderators often rely on social signals such as whether or not a user has an account or a profile page as clues that users may cause problems. Reliance on these clues can lead to overprofiling bias when moderators focus on these signals but overlook the misbehavior of others. We propose that algorithmic flagging systems deployed to improve the efficiency of moderation work can also make moderation actions more fair to these users by reducing reliance on social signals and making norm violations by everyone else more visible. We analyze moderator behavior in Wikipedia as mediated by RCFilters, a system which displays social signals and algorithmic flags, and estimate the causal effect of being flagged on moderator actions. We show that algorithmically flagged edits are reverted more often, especially those by established editors with positive social signals, and that flagging decreases the likelihood that moderation actions will be undone. Our results suggest that algorithmic flagging systems can lead to increased fairness in some contexts but that the relationship is complex and contingent. △ Less

Submitted 5 April, 2021; v1 submitted 4 June, 2020; originally announced June 2020.

Comments: 27 pages, 11 figures, ACM CSCW

ACM Class: K.4.3

Journal ref: Proc. ACM Hum.-Comput. Interact. 5, CSCW1, Article 56 (April 2021), 27 pages

How individual behaviors drive inequality in online community sizes: an agent-based simulation

Authors: Jeremy Foote, Nathan TeBlunthuis, Benjamin Mako Hill, Aaron Shaw

Abstract: Why are online community sizes so extremely unequal? Most answers to this question have pointed to general mathematical processes drawn from physics like cumulative advantage. These explanations provide little insight into specific social dynamics or decisions that individuals make when joining and leaving communities. In addition, explanations in terms of cumulative advantage do not draw from the… ▽ More Why are online community sizes so extremely unequal? Most answers to this question have pointed to general mathematical processes drawn from physics like cumulative advantage. These explanations provide little insight into specific social dynamics or decisions that individuals make when joining and leaving communities. In addition, explanations in terms of cumulative advantage do not draw from the enormous body of social computing research that studies individual behavior. Our work bridges this divide by testing whether two influential social mechanisms used to explain community joining can also explain the distribution of community sizes. Using agent-based simulations, we evaluate how well individual-level processes of social exposure and decisions based on individual expected benefits reproduce empirical community size data from Reddit. Our simulations contribute to social computing theory by providing evidence that both processes together---but neither alone---generate realistic distributions of community sizes. Our results also illustrate the potential value of agent-based simulation to online community researchers to both evaluate and bridge individual and group-level theories. △ Less

Submitted 4 June, 2020; originally announced June 2020.

ACM Class: K.4.3

Stochastic Security: Adversarial Defense Using Long-Run Dynamics of Energy-Based Models

Authors: Mitch Hill, Jonathan Mitchell, Song-Chun Zhu

Abstract: The vulnerability of deep networks to adversarial attacks is a central problem for deep learning from the perspective of both cognition and security. The current most successful defense method is to train a classifier using adversarial images created during learning. Another defense approach involves transformation or purification of the original input to remove adversarial signals before the imag… ▽ More The vulnerability of deep networks to adversarial attacks is a central problem for deep learning from the perspective of both cognition and security. The current most successful defense method is to train a classifier using adversarial images created during learning. Another defense approach involves transformation or purification of the original input to remove adversarial signals before the image is classified. We focus on defending naturally-trained classifiers using Markov Chain Monte Carlo (MCMC) sampling with an Energy-Based Model (EBM) for adversarial purification. In contrast to adversarial training, our approach is intended to secure pre-existing and highly vulnerable classifiers. The memoryless behavior of long-run MCMC sampling will eventually remove adversarial signals, while metastable behavior preserves consistent appearance of MCMC samples after many steps to allow accurate long-run prediction. Balancing these factors can lead to effective purification and robust classification. We evaluate adversarial defense with an EBM using the strongest known attacks against purification. Our contributions are 1) an improved method for training EBM's with realistic long-run MCMC samples, 2) an Expectation-Over-Transformation (EOT) defense that resolves theoretical ambiguities for stochastic defenses and from which the EOT attack naturally follows, and 3) state-of-the-art adversarial defense for naturally-trained classifiers and competitive defense compared to adversarially-trained classifiers on Cifar-10, SVHN, and Cifar-100. Code and pre-trained models are available at https://github.com/point0bar1/ebm-defense. △ Less

Submitted 18 March, 2021; v1 submitted 27 May, 2020; originally announced May 2020.

Comments: ICLR 2021

Nanotechnology-inspired Information Processing Systems of the Future

Authors: Randy Bryant, Mark Hill, Tom Kazior, Daniel Lee, Jie Liu, Klara Nahrstedt, Vijay Narayanan, Jan Rabaey, Hava Siegelmann, Naresh Shanbhag, Naveen Verma, H. -S. Philip Wong

Abstract: Nanoscale semiconductor technology has been a key enabler of the computing revolution. It has done so via advances in new materials and manufacturing processes that resulted in the size of the basic building block of computing systems - the logic switch and memory devices - being reduced into the nanoscale regime. Nanotechnology has provided increased computing functionality per unit volume, energ… ▽ More Nanoscale semiconductor technology has been a key enabler of the computing revolution. It has done so via advances in new materials and manufacturing processes that resulted in the size of the basic building block of computing systems - the logic switch and memory devices - being reduced into the nanoscale regime. Nanotechnology has provided increased computing functionality per unit volume, energy, and cost. In order for computing systems to continue to deliver substantial benefits for the foreseeable future to society at large, it is critical that the very notion of computing be examined in the light of nanoscale realities. In particular, one needs to ask what it means to compute when the very building block - the logic switch - no longer exhibits the level of determinism required by the von Neumann architecture. There needs to be a sustained and heavy investment in a nation-wide Vertically Integrated Semiconductor Ecosystem (VISE). VISE is a program in which research and development is conducted seamlessly across the entire compute stack - from applications, systems and algorithms, architectures, circuits and nanodevices, and materials. A nation-wide VISE provides clear strategic advantages in ensuring the US's global superiority in semiconductors. First, a VISE provides the highest quality seed-corn for nurturing transformative ideas that are critically needed today in order for nanotechnology-inspired computing to flourish. It does so by dramatically opening up new areas of semiconductor research that are inspired and driven by new application needs. Second, a VISE creates a very high barrier to entry from foreign competitors because it is extremely hard to establish, and even harder to duplicate. △ Less

Submitted 5 May, 2020; originally announced May 2020.

Comments: A Computing Community Consortium (CCC) workshop report, 18 pages

Report number: ccc2016report_3

BARR-C:2018 and MISRA C:2012: Synergy Between the Two Most Widely Used C Coding Standards

Authors: Roberto Bagnara, Michael Barr, Patricia M. Hill

Abstract: The Barr Group's Embedded C Coding Standard (BARR-C:2018, which originates from the 2009 Netrino's Embedded C Coding Standard) is, for coding standards used by the embedded system industry, second only in popularity to MISRA C. However, the choice between MISRA C:2012 and BARR-C:2018 needs not be a hard decision since they are complementary in two quite different ways. On the one hand, BARR-C:2018… ▽ More The Barr Group's Embedded C Coding Standard (BARR-C:2018, which originates from the 2009 Netrino's Embedded C Coding Standard) is, for coding standards used by the embedded system industry, second only in popularity to MISRA C. However, the choice between MISRA C:2012 and BARR-C:2018 needs not be a hard decision since they are complementary in two quite different ways. On the one hand, BARR-C:2018 has removed all the incompatibilities with respect to MISRA C:2012 that were present in the previous edition (BARR-C:2013). As a result, disregarding programming style, BARR-C:2018 defines a subset of C that, while preventing a significant number of programming errors, is larger than the one defined by MISRA C:2012. On the other hand, concerning programming style, whereas MISRA C leaves this to individual organizations, BARR-C:2018 defines a programming style aimed primarily at minimizing programming errors. As a result, BARR-C:2018 can be seen as a first, dramatically useful step to C language subsetting that is suitable for all kinds of projects; critical projects can then evolve toward MISRA C:2012 compliance smoothly while maintaining the BARR-C programming style. In this paper, we introduce BARR-C:2018, we describe its relationship with MISRA C:2012, and we discuss the parallel and serial adoption of the two coding standards. △ Less

Submitted 15 March, 2020; originally announced March 2020.

Comments: 14 pages, 1 figure

MSC Class: 68N15 ACM Class: F.3.1; D.3.0

Single Unit Status in Deep Convolutional Neural Network Codes for Face Identification: Sparseness Redefined

Authors: Connor J. Parde, Y. Ivette Colón, Matthew Q. Hill, Carlos D. Castillo, Prithviraj Dhar, Alice J. O'Toole

Abstract: Deep convolutional neural networks (DCNNs) trained for face identification develop representations that generalize over variable images, while retaining subject (e.g., gender) and image (e.g., viewpoint) information. Identity, gender, and viewpoint codes were studied at the "neural unit" and ensemble levels of a face-identification network. At the unit level, identification, gender classification,… ▽ More Deep convolutional neural networks (DCNNs) trained for face identification develop representations that generalize over variable images, while retaining subject (e.g., gender) and image (e.g., viewpoint) information. Identity, gender, and viewpoint codes were studied at the "neural unit" and ensemble levels of a face-identification network. At the unit level, identification, gender classification, and viewpoint estimation were measured by deleting units to create variably-sized, randomly-sampled subspaces at the top network layer. Identification of 3,531 identities remained high (area under the ROC approximately 1.0) as dimensionality decreased from 512 units to 16 (0.95), 4 (0.80), and 2 (0.72) units. Individual identities separated statistically on every top-layer unit. Cross-unit responses were minimally correlated, indicating that units code non-redundant identity cues. This "distributed" code requires only a sparse, random sample of units to identify faces accurately. Gender classification declined gradually and viewpoint estimation fell steeply as dimensionality decreased. Individual units were weakly predictive of gender and viewpoint, but ensembles proved effective predictors. Therefore, distributed and sparse codes co-exist in the network units to represent different face attributes. At the ensemble level, principal component analysis of face representations showed that identity, gender, and viewpoint information separated into high-dimensional subspaces, ordered by explained variance. Identity, gender, and viewpoint information contributed to all individual unit responses, undercutting a neural tuning analogy for face attributes. Interpretation of neural-like codes from DCNNs, and by analogy, high-level visual codes, cannot be inferred from single unit responses. Instead, "meaning" is encoded by directions in the high-dimensional space. △ Less

Submitted 1 March, 2020; v1 submitted 14 February, 2020; originally announced February 2020.

Thermodynamic Computing

Authors: Tom Conte, Erik DeBenedictis, Natesh Ganesh, Todd Hylton, John Paul Strachan, R. Stanley Williams, Alexander Alemi, Lee Altenberg, Gavin Crooks, James Crutchfield, Lidia del Rio, Josh Deutsch, Michael DeWeese, Khari Douglas, Massimiliano Esposito, Michael Frank, Robert Fry, Peter Harsha, Mark Hill, Christopher Kello, Jeff Krichmar, Suhas Kumar, Shih-Chii Liu, Seth Lloyd, Matteo Marsili , et al. (14 additional authors not shown)

Abstract: The hardware and software foundations laid in the first half of the 20th Century enabled the computing technologies that have transformed the world, but these foundations are now under siege. The current computing paradigm, which is the foundation of much of the current standards of living that we now enjoy, faces fundamental limitations that are evident from several perspectives. In terms of hard… ▽ More The hardware and software foundations laid in the first half of the 20th Century enabled the computing technologies that have transformed the world, but these foundations are now under siege. The current computing paradigm, which is the foundation of much of the current standards of living that we now enjoy, faces fundamental limitations that are evident from several perspectives. In terms of hardware, devices have become so small that we are struggling to eliminate the effects of thermodynamic fluctuations, which are unavoidable at the nanometer scale. In terms of software, our ability to imagine and program effective computational abstractions and implementations are clearly challenged in complex domains. In terms of systems, currently five percent of the power generated in the US is used to run computing systems - this astonishing figure is neither ecologically sustainable nor economically scalable. Economically, the cost of building next-generation semiconductor fabrication plants has soared past $10 billion. All of these difficulties - device scaling, software complexity, adaptability, energy consumption, and fabrication economics - indicate that the current computing paradigm has matured and that continued improvements along this path will be limited. If technological progress is to continue and corresponding social and economic benefits are to continue to accrue, computing must become much more capable, energy efficient, and affordable. We propose that progress in computing can continue under a united, physically grounded, computational paradigm centered on thermodynamics. Herein we propose a research agenda to extend these thermodynamic foundations into complex, non-equilibrium, self-organizing systems and apply them holistically to future computing systems that will harness nature's innate computational capacity. We call this type of computing "Thermodynamic Computing" or TC. △ Less

Submitted 14 November, 2019; v1 submitted 5 November, 2019; originally announced November 2019.

Comments: A Computing Community Consortium (CCC) workshop report, 36 pages

Report number: ccc2019report_6

A Forensic Qualitative Analysis of Contributions to Wikipedia from Anonymity Seeking Users

Authors: Kaylea Champion, Nora McDonald, Stephanie Bankes, Joseph Zhang, Rachel Greenstadt, Andrea Forte, Benjamin Mako Hill

Abstract: By choice or by necessity, some contributors to commons-based peer production sites use privacy-protecting services to remain anonymous. As anonymity seekers, users of the Tor network have been cast both as ill-intentioned vandals and as vulnerable populations concerned with their privacy. In this study, we use a dataset drawn from a corpus of Tor edits to Wikipedia to uncover the character of Tor… ▽ More By choice or by necessity, some contributors to commons-based peer production sites use privacy-protecting services to remain anonymous. As anonymity seekers, users of the Tor network have been cast both as ill-intentioned vandals and as vulnerable populations concerned with their privacy. In this study, we use a dataset drawn from a corpus of Tor edits to Wikipedia to uncover the character of Tor users' contributions. We build in-depth narrative descriptions of Tor users' actions and conduct a thematic analysis that places their editing activity into seven broad groups. We find that although their use of a privacy-protecting service marks them as unusual within Wikipedia, the character of many Tor users' contributions is in line with the expectations and norms of Wikipedia. However, our themes point to several important places where lack of trust promotes disorder, and to contributions where risks to contributors, service providers, and communities are unaligned. △ Less

Submitted 17 September, 2019; originally announced September 2019.

Journal ref: Proc. ACM Hum.-Comput. Interact. 3, CSCW, Article 53 (November 2019)

MOD: Minimally Ordered Durable Datastructures for Persistent Memory

Authors: Swapnil Haria, Mark D. Hill, Michael M. Swift

Abstract: Persistent Memory (PM) makes possible recoverable applications that can preserve application progress across system reboots and power failures. Actual recoverability requires careful ordering of cacheline flushes, currently done in two extreme ways. On one hand, expert programmers have reasoned deeply about consistency and durability to create applications centered on a single custom-crafted durab… ▽ More Persistent Memory (PM) makes possible recoverable applications that can preserve application progress across system reboots and power failures. Actual recoverability requires careful ordering of cacheline flushes, currently done in two extreme ways. On one hand, expert programmers have reasoned deeply about consistency and durability to create applications centered on a single custom-crafted durable datastructure. On the other hand, less-expert programmers have used software transaction memory (STM) to make atomic one or more updates, albeit at a significant performance cost due largely to ordered log updates. In this work, we propose the middle ground of composable persistent datastructures called Minimally Ordered Durable (MOD) datastructures. MOD is a C++ library of several datastructures---currently, map, set, stack, queue and vector--- that often perform better than STM and yet are relatively easy to use. They allow multiple updates to one or more datastructures to be atomic with respect to failure. Moreover, we provide a recipe to create more recoverable datastructures. MOD is motivated by our analysis of real Intel Optane PM hardware showing that allowing unordered, overlapping flushes significantly improves performance. MOD reduces ordering by adapting existing techniques for out-of-place updates (like shadow paging) with space-reducing structural sharing (from functional programming). MOD exposes a Basic interface for single updates and a Composition interface for atomically performing multiple updates. Relative to the state-of-the-art Intel PMDK v1.5 STM, MOD improves map, set, stack, queue microbenchmark performance by 40%, and speeds up application benchmark performance by 38%. △ Less

Submitted 21 August, 2019; originally announced August 2019.

Comments: 13 pages

P2L: Predicting Transfer Learning for Images and Semantic Relations

Authors: Bishwaranjan Bhattacharjee, John R. Kender, Matthew Hill, Parijat Dube, Siyu Huo, Michael R. Glass, Brian Belgodere, Sharath Pankanti, Noel Codella, Patrick Watson

Abstract: Transfer learning enhances learning across tasks, by leveraging previously learned representations -- if they are properly chosen. We describe an efficient method to accurately estimate the appropriateness of a previously trained model for use in a new learning task. We use this measure, which we call "Predict To Learn" ("P2L"), in the two very different domains of images and semantic relations, w… ▽ More Transfer learning enhances learning across tasks, by leveraging previously learned representations -- if they are properly chosen. We describe an efficient method to accurately estimate the appropriateness of a previously trained model for use in a new learning task. We use this measure, which we call "Predict To Learn" ("P2L"), in the two very different domains of images and semantic relations, where it predicts, from a set of "source" models, the one model most likely to produce effective transfer for training a given "target" model. We validate our approach thoroughly, by assembling a collection of candidate source models, then fine-tuning each candidate to perform each of a collection of target tasks, and finally measuring how well transfer has been enhanced. Across 95 tasks within multiple domains (images classification and semantic relations), the P2L approach was able to select the best transfer learning model on average, while the heuristic of choosing model trained with the largest data set selected the best model in only 55 cases. These results suggest that P2L captures important information in common between source and target tasks, and that this shared informational structure contributes to successful transfer learning more than simple data size. △ Less

Submitted 15 October, 2020; v1 submitted 20 August, 2019; originally announced August 2019.

Comments: 10 pages, 8 figures, 4 tables

Accelerator-level Parallelism

Authors: Mark D. Hill, Vijay Janapa Reddi

Abstract: Future applications demand more performance, but technology advances have been faltering. A promising approach to further improve computer system performance under energy constraints is to employ hardware accelerators. Already today, mobile systems concurrently employ multiple accelerators in what we call accelerator-level parallelism (ALP). To spread the benefits of ALP more broadly, we charge co… ▽ More Future applications demand more performance, but technology advances have been faltering. A promising approach to further improve computer system performance under energy constraints is to employ hardware accelerators. Already today, mobile systems concurrently employ multiple accelerators in what we call accelerator-level parallelism (ALP). To spread the benefits of ALP more broadly, we charge computer scientists to develop the science needed to best achieve the performance and cost goals of ALP hardware and software. △ Less

Submitted 24 November, 2021; v1 submitted 2 July, 2019; originally announced July 2019.

Comments: 6 pages, 3 figures, & 7 references

Don't Persist All : Efficient Persistent Data Structures

Authors: Pratyush Mahapatra, Mark D. Hill, Michael M. Swift

Abstract: Data structures used in software development have inbuilt redundancy to improve software reliability and to speed up performance. Examples include a Doubly Linked List which allows a faster deletion due to the presence of the previous pointer. With the introduction of Persistent Memory, storing the redundant data fields into persistent memory adds a significant write overhead, and reduces performa… ▽ More Data structures used in software development have inbuilt redundancy to improve software reliability and to speed up performance. Examples include a Doubly Linked List which allows a faster deletion due to the presence of the previous pointer. With the introduction of Persistent Memory, storing the redundant data fields into persistent memory adds a significant write overhead, and reduces performance. In this work, we focus on three data structures - Doubly Linked List, B+Tree and Hashmap, and showcase alternate partly persistent implementations where we only store a limited set of data fields to persistent memory. After a crash/restart, we use the persistent data fields to recreate the data structures along with the redundant data fields. We compare our implementation with the base implementation and show that we achieve speedups around 5-20% for some data structures, and up to 165% for a flush-dominated data structure. △ Less

Submitted 29 May, 2019; originally announced May 2019.

Comments: 10 pages, 12 figures

Learning Non-Convergent Non-Persistent Short-Run MCMC Toward Energy-Based Model

Authors: Erik Nijkamp, Mitch Hill, Song-Chun Zhu, Ying Nian Wu

Abstract: This paper studies a curious phenomenon in learning energy-based model (EBM) using MCMC. In each learning iteration, we generate synthesized examples by running a non-convergent, non-mixing, and non-persistent short-run MCMC toward the current model, always starting from the same initial distribution such as uniform noise distribution, and always running a fixed number of MCMC steps. After generat… ▽ More This paper studies a curious phenomenon in learning energy-based model (EBM) using MCMC. In each learning iteration, we generate synthesized examples by running a non-convergent, non-mixing, and non-persistent short-run MCMC toward the current model, always starting from the same initial distribution such as uniform noise distribution, and always running a fixed number of MCMC steps. After generating synthesized examples, we then update the model parameters according to the maximum likelihood learning gradient, as if the synthesized examples are fair samples from the current model. We treat this non-convergent short-run MCMC as a learned generator model or a flow model. We provide arguments for treating the learned non-convergent short-run MCMC as a valid model. We show that the learned short-run MCMC is capable of generating realistic images. More interestingly, unlike traditional EBM or MCMC, the learned short-run MCMC is capable of reconstructing observed images and interpolating between images, like generator or flow models. The code can be found in the Appendix. △ Less

Submitted 25 November, 2019; v1 submitted 22 April, 2019; originally announced April 2019.

Are anonymity-seekers just like everybody else? An analysis of contributions to Wikipedia from Tor

Authors: Chau Tran, Kaylea Champion, Andrea Forte, Benjamin Mako Hill, Rachel Greenstadt

Abstract: User-generated content sites routinely block contributions from users of privacy-enhancing proxies like Tor because of a perception that proxies are a source of vandalism, spam, and abuse. Although these blocks might be effective, collateral damage in the form of unrealized valuable contributions from anonymity seekers is invisible. One of the largest and most important user-generated content site… ▽ More User-generated content sites routinely block contributions from users of privacy-enhancing proxies like Tor because of a perception that proxies are a source of vandalism, spam, and abuse. Although these blocks might be effective, collateral damage in the form of unrealized valuable contributions from anonymity seekers is invisible. One of the largest and most important user-generated content sites, Wikipedia, has attempted to block contributions from Tor users since as early as 2005. We demonstrate that these blocks have been imperfect and that thousands of attempts to edit on Wikipedia through Tor have been successful. We draw upon several data sources and analytical techniques to measure and describe the history of Tor editing on Wikipedia over time and to compare contributions from Tor users to those from other groups of Wikipedia users. Our analysis suggests that although Tor users who slip through Wikipedia's ban contribute content that is more likely to be reverted and to revert others, their contributions are otherwise similar in quality to those from other unregistered participants and to the initial contributions of registered users. △ Less

Submitted 15 February, 2020; v1 submitted 8 April, 2019; originally announced April 2019.

Comments: To appear in the IEEE Symposium on Security & Privacy, May 2020

On the Anatomy of MCMC-Based Maximum Likelihood Learning of Energy-Based Models

Authors: Erik Nijkamp, Mitch Hill, Tian Han, Song-Chun Zhu, Ying Nian Wu

Abstract: This study investigates the effects of Markov chain Monte Carlo (MCMC) sampling in unsupervised Maximum Likelihood (ML) learning. Our attention is restricted to the family of unnormalized probability densities for which the negative log density (or energy function) is a ConvNet. We find that many of the techniques used to stabilize training in previous studies are not necessary. ML learning with a… ▽ More This study investigates the effects of Markov chain Monte Carlo (MCMC) sampling in unsupervised Maximum Likelihood (ML) learning. Our attention is restricted to the family of unnormalized probability densities for which the negative log density (or energy function) is a ConvNet. We find that many of the techniques used to stabilize training in previous studies are not necessary. ML learning with a ConvNet potential requires only a few hyper-parameters and no regularization. Using this minimal framework, we identify a variety of ML learning outcomes that depend solely on the implementation of MCMC sampling. On one hand, we show that it is easy to train an energy-based model which can sample realistic images with short-run Langevin. ML can be effective and stable even when MCMC samples have much higher energy than true steady-state samples throughout training. Based on this insight, we introduce an ML method with purely noise-initialized MCMC, high-quality short-run synthesis, and the same budget as ML with informative MCMC initialization such as CD or PCD. Unlike previous models, our energy model can obtain realistic high-diversity samples from a noise signal after training. On the other hand, ConvNet potentials learned with non-convergent MCMC do not have a valid steady-state and cannot be considered approximate unnormalized densities of the training data because long-run MCMC samples differ greatly from observed images. We show that it is much harder to train a ConvNet potential to learn a steady-state over realistic images. To our knowledge, long-run MCMC samples of all previous models lose the realism of short-run samples. With correct tuning of Langevin noise, we train the first ConvNet potentials for which long-run and steady-state MCMC samples are realistic images. △ Less

Submitted 27 November, 2019; v1 submitted 29 March, 2019; originally announced March 2019.

Comments: Code available at: https://github.com/point0bar1/ebm-anatomy

Journal ref: AAAI 2020

Three Other Models of Computer System Performance

Authors: Mark D. Hill

Abstract: This note argues for more use of simple models beyond Amdahl's Law: Bottleneck Analysis, Little's Law, and a M/M/1 Queue. This note argues for more use of simple models beyond Amdahl's Law: Bottleneck Analysis, Little's Law, and a M/M/1 Queue. △ Less

Submitted 9 January, 2019; originally announced January 2019.

Comments: 4 pages + references; 4 figures

Divergence Triangle for Joint Training of Generator Model, Energy-based Model, and Inference Model

Authors: Tian Han, Erik Nijkamp, Xiaolin Fang, Mitch Hill, Song-Chun Zhu, Ying Nian Wu

Abstract: This paper proposes the divergence triangle as a framework for joint training of generator model, energy-based model and inference model. The divergence triangle is a compact and symmetric (anti-symmetric) objective function that seamlessly integrates variational learning, adversarial learning, wake-sleep algorithm, and contrastive divergence in a unified probabilistic formulation. This unificatio… ▽ More This paper proposes the divergence triangle as a framework for joint training of generator model, energy-based model and inference model. The divergence triangle is a compact and symmetric (anti-symmetric) objective function that seamlessly integrates variational learning, adversarial learning, wake-sleep algorithm, and contrastive divergence in a unified probabilistic formulation. This unification makes the processes of sampling, inference, energy evaluation readily available without the need for costly Markov chain Monte Carlo methods. Our experiments demonstrate that the divergence triangle is capable of learning (1) an energy-based model with well-formed energy landscape, (2) direct sampling in the form of a generator network, and (3) feed-forward inference that faithfully reconstructs observed as well as synthesized data. The divergence triangle is a robust training method that can learn from incomplete data. △ Less

Submitted 31 January, 2019; v1 submitted 28 December, 2018; originally announced December 2018.

Deep Convolutional Neural Networks in the Face of Caricature: Identity and Image Revealed

Authors: Matthew Q. Hill, Connor J. Parde, Carlos D. Castillo, Y. Ivette Colon, Rajeev Ranjan, Jun-Cheng Chen, Volker Blanz, Alice J. O'Toole

Abstract: Real-world face recognition requires an ability to perceive the unique features of an individual face across multiple, variable images. The primate visual system solves the problem of image invariance using cascades of neurons that convert images of faces into categorical representations of facial identity. Deep convolutional neural networks (DCNNs) also create generalizable face representations,… ▽ More Real-world face recognition requires an ability to perceive the unique features of an individual face across multiple, variable images. The primate visual system solves the problem of image invariance using cascades of neurons that convert images of faces into categorical representations of facial identity. Deep convolutional neural networks (DCNNs) also create generalizable face representations, but with cascades of simulated neurons. DCNN representations can be examined in a multidimensional "face space", with identities and image parameters quantified via their projections onto the axes that define the space. We examined the organization of viewpoint, illumination, gender, and identity in this space. We show that the network creates a highly organized, hierarchically nested, face similarity structure in which information about face identity and imaging characteristics coexist. Natural image variation is accommodated in this hierarchy, with face identity nested under gender, illumination nested under identity, and viewpoint nested under illumination. To examine identity, we caricatured faces and found that network identification accuracy increased with caricature level, and--mimicking human perception--a caricatured distortion of a face "resembled" its veridical counterpart. Caricatures improved performance by moving the identity away from other identities in the face space and minimizing the effects of illumination and viewpoint. Deep networks produce face representations that solve long-standing computational problems in generalized face recognition. They also provide a unitary theoretical framework for reconciling decades of behavioral and neural results that emphasized either the image or the object/face in representations, without understanding how a neural code could seamlessly accommodate both. △ Less

Submitted 28 December, 2018; originally announced December 2018.

Comments: 8 pages, 5 figures

The MISRA C Coding Standard and its Role in the Development and Analysis of Safety- and Security-Critical Embedded Software

Authors: Roberto Bagnara, Abramo Bagnara, Patricia M. Hill

Abstract: The MISRA project started in 1990 with the mission of providing world-leading best practice guidelines for the safe and secure application of both embedded control systems and standalone software. MISRA C is a coding standard defining a subset of the C language, initially targeted at the automotive sector, but now adopted across all industry sectors that develop C software in safety- and/or securi… ▽ More The MISRA project started in 1990 with the mission of providing world-leading best practice guidelines for the safe and secure application of both embedded control systems and standalone software. MISRA C is a coding standard defining a subset of the C language, initially targeted at the automotive sector, but now adopted across all industry sectors that develop C software in safety- and/or security-critical contexts. In this paper, we introduce MISRA C, its role in the development of critical software, especially in embedded systems, its relevance to industry safety standards, as well as the challenges of working with a general-purpose programming language standard that is written in natural language with a slow evolution over the last 40+ years. We also outline the role of static analysis in the automatic checking of compliance with respect to MISRA C, and the role of the MISRA C language subset in enabling a wider application of formal methods to industrial software written in C. △ Less

Submitted 4 September, 2018; originally announced September 2018.

Comments: 19 pages, 1 figure, 2 tables

MSC Class: 68N15 ACM Class: F.3.1; D.3.0

Improving Transferability of Deep Neural Networks

Authors: Parijat Dube, Bishwaranjan Bhattacharjee, Elisabeth Petit-Bois, Matthew Hill

Abstract: Learning from small amounts of labeled data is a challenge in the area of deep learning. This is currently addressed by Transfer Learning where one learns the small data set as a transfer task from a larger source dataset. Transfer Learning can deliver higher accuracy if the hyperparameters and source dataset are chosen well. One of the important parameters is the learning rate for the layers of t… ▽ More Learning from small amounts of labeled data is a challenge in the area of deep learning. This is currently addressed by Transfer Learning where one learns the small data set as a transfer task from a larger source dataset. Transfer Learning can deliver higher accuracy if the hyperparameters and source dataset are chosen well. One of the important parameters is the learning rate for the layers of the neural network. We show through experiments on the ImageNet22k and Oxford Flowers datasets that improvements in accuracy in range of 127% can be obtained by proper choice of learning rates. We also show that the images/label parameter for a dataset can potentially be used to determine optimal learning rates for the layers to get the best overall accuracy. We additionally validate this method on a sample of real-world image classification tasks from a public visual recognition API. △ Less

Submitted 30 July, 2018; originally announced July 2018.

Comments: 15 pages, 11 figures, 2 tables, Workshop on Domain Adaptation for Visual Understanding (Joint IJCAI/ECAI/AAMAS/ICML 2018 Workshop) Keywords: deep learning, transfer learning, finetuning, deep neural network, experimental

Testing Deep Neural Networks

Authors: Youcheng Sun, Xiaowei Huang, Daniel Kroening, James Sharp, Matthew Hill, Rob Ashmore

Abstract: Deep neural networks (DNNs) have a wide range of applications, and software employing them must be thoroughly tested, especially in safety-critical domains. However, traditional software test coverage metrics cannot be applied directly to DNNs. In this paper, inspired by the MC/DC coverage criterion, we propose a family of four novel test criteria that are tailored to structural features of DNNs a… ▽ More Deep neural networks (DNNs) have a wide range of applications, and software employing them must be thoroughly tested, especially in safety-critical domains. However, traditional software test coverage metrics cannot be applied directly to DNNs. In this paper, inspired by the MC/DC coverage criterion, we propose a family of four novel test criteria that are tailored to structural features of DNNs and their semantics. We validate the criteria by demonstrating that the generated test inputs guided via our proposed coverage criteria are able to capture undesired behaviours in a DNN. Test cases are generated using a symbolic approach and a gradient-based heuristic search. By comparing them with existing methods, we show that our criteria achieve a balance between their ability to find bugs (proxied using adversarial examples) and the computational cost of test case generation. Our experiments are conducted on state-of-the-art DNNs obtained using popular open source datasets, including MNIST, CIFAR-10 and ImageNet. △ Less

Submitted 15 April, 2019; v1 submitted 10 March, 2018; originally announced March 2018.