Layer Ensemble Averaging for Improving Memristor-Based Artificial Neural Network Performance
Authors:
Osama Yousuf,
Brian Hoskins,
Karthick Ramu,
Mitchell Fream,
William A. Borders,
Advait Madhavan,
Matthew W. Daniels,
Andrew Dienstfrey,
Jabez J. McClelland,
Martin Lueker-Boden,
Gina C. Adam
Abstract:
Artificial neural networks have advanced due to scaling dimensions, but conventional computing faces inefficiency due to the von Neumann bottleneck. In-memory computation architectures, like memristors, offer promise but face challenges due to hardware non-idealities. This work proposes and experimentally demonstrates layer ensemble averaging, a technique to map pre-trained neural network solution…
▽ More
Artificial neural networks have advanced due to scaling dimensions, but conventional computing faces inefficiency due to the von Neumann bottleneck. In-memory computation architectures, like memristors, offer promise but face challenges due to hardware non-idealities. This work proposes and experimentally demonstrates layer ensemble averaging, a technique to map pre-trained neural network solutions from software to defective hardware crossbars of emerging memory devices and reliably attain near-software performance on inference. The approach is investigated using a custom 20,000-device hardware prototyping platform on a continual learning problem where a network must learn new tasks without catastrophically forgetting previously learned information. Results demonstrate that by trading off the number of devices required for layer mapping, layer ensemble averaging can reliably boost defective memristive network performance up to the software baseline. For the investigated problem, the average multi-task classification accuracy improves from 61 % to 72 % (< 1 % of software baseline) using the proposed approach.
△ Less
Submitted 23 April, 2024;
originally announced April 2024.
A 3.3 Gbps SPAD-Based Quantum Random Number Generator
Authors:
Pouyan Keshavarzian,
Karthick Ramu,
Duy Tang,
Carlos Weill,
Francesco Gramuglia,
Shyue Seng Tan,
Michelle Tng,
Louis Lim,
Elgin Quek,
Denis Mandich,
Mario Stipčević,
Edoardo Charbon
Abstract:
Quantum random number generators are a burgeoning technology used for a variety of applications, including modern security and encryption systems. Typical methods exploit an entropy source combined with an extraction or bit generation circuit in order to produce a random string. In integrated designs there is often little modelling or analytical description of the entropy source, circuit extractio…
▽ More
Quantum random number generators are a burgeoning technology used for a variety of applications, including modern security and encryption systems. Typical methods exploit an entropy source combined with an extraction or bit generation circuit in order to produce a random string. In integrated designs there is often little modelling or analytical description of the entropy source, circuit extraction and post-processing provided. In this work, we first discuss theory on the quantum random flip-flop (QRFF), which elucidates the role of circuit imperfections that manifest themselves in bias and correlation. Then, a Verilog-AMS model is developed in order to validate the analytical model in simulation. A novel transistor implementation of the QRFF circuit is presented, which enables compensation of the degradation in entropy inherent to the finite non-symmetric transitions of the random flip-flop. Finally, a full system containing two independent arrays of the QRFF circuit is manufactured and tested in a 55 nm Bipolar-CMOS-DMOS (BCD) technology node, demonstrating bit generation statistics that are commensurate to the developed model. The full chip is able to generate 3.3 Gbps of data when operated with an external LED, whereas an individual QRFF can generate 25 Mbps each of random data while maintaining a Shannon entropy bound > 0.997, which is one of the highest per pixel bit generation rates to date. NIST STS is used to benchmark the generated bit strings, thereby validating the QRFF circuit as an excellent candidate for fully-integrated QRNGs.
△ Less
Submitted 11 September, 2022;
originally announced September 2022.
