Workshop
Machine Learning with New Compute Paradigms
Jannes Gladrow 路 Benjamin Scellier 路 Eric Xing 路 Babak Rahmani 路 Francesca Parmigiani 路 Paul Prucnal 路 Cheng Zhang
Room 235 - 236
Sat 16 Dec, 7 a.m. PST
As GPU computing comes closer to a plateau in terms of efficiency and cost due to Moore鈥檚 law reaching its limit, there is a growing need to explore alternative computing paradigms, such as (opto-)analog, neuromorphic, and low-power computing. This NeurIPS workshop aims to unite researchers from machine learning and alternative computation fields to establish a new hardware-ML feedback loop.By co-designing models with specialized accelerators, we can leverage the benefits of increased throughput or lower per-flop power consumption. Novel devices hold the potential to further accelerate standard deep learning or even enable efficient inference and training of hitherto compute-constrained model classes. However, new compute paradigms typically present challenges such as intrinsic noise, restricted sets of compute operations, or limited bit-depth, and thus require model-hardware co-design. This workshop鈥檚 goal is to foster cross-disciplinary collaboration to capitalize on the opportunities offered by emerging AI accelerators.
Schedule
Sat 7:00 a.m. - 7:10 a.m.
|
Opening Remarks
(
Talk
)
>
|
Cheng Zhang 馃敆 |
Sat 7:10 a.m. - 7:30 a.m.
|
Computing with physical systems: reimagining special-purpose computers from the bottom up
(
Invited Talk
)
>
SlidesLive Video |
Peter McMahon 馃敆 |
Sat 7:30 a.m. - 7:40 a.m.
|
SpiNNaker2: A Large-Scale Neuromorphic System for Event-Based and Asynchronous Machine Learning
(
Oral
)
>
link
SlidesLive Video |
14 presentersHector Gonzalez 路 Jiaxin Huang 路 Florian Kelber 路 Khaleelulla Khan Nazeer 路 Tim Hauke Langer 路 Chen Liu 路 Matthias Lohrmann 路 Amirhossein Rostami 路 Mark Schoene 路 Bernhard Vogginger 路 Timo Wunderlich 路 Yexin Yan 路 Mahmoud Akl 路 Christian Mayr |
Sat 7:40 a.m. - 7:50 a.m.
|
Scaling-up Memristor Monte Carlo with magnetic domain-wall physics
(
Oral
)
>
link
SlidesLive Video |
11 presentersThomas Dalgaty 路 Shogo Yamada 路 Anca Molnos 路 Eiji Kawasaki 路 Thomas Mesquida 路 Rummens Fran莽ois 路 TATSUO SHIBATA 路 Yukihiro Urakawa 路 Yukio Terasaki 路 Tomoyuki Sasaki 路 Marc Duranton |
Sat 7:50 a.m. - 8:10 a.m.
|
Analog AI Accelerators
(
Invited Talk
)
>
link
SlidesLive Video |
Jesus del Alamo 馃敆 |
Sat 8:10 a.m. - 8:25 a.m.
|
Break
|
馃敆 |
Sat 8:25 a.m. - 8:35 a.m.
|
Analog-Optical Computation for optimization and machine-learning inference
(
Talk
)
>
link
SlidesLive Video |
Jannes Gladrow 馃敆 |
Sat 8:35 a.m. - 8:45 a.m.
|
Bayesian Metaplasticity from Synaptic Uncertainty
(
Oral
)
>
link
SlidesLive Video |
Djohan Bonnet 路 Tifenn HIRTZLIN 路 Tarcisius Januel 路 Thomas Dalgaty 路 Damien Querlioz 路 Elisa Vianello 馃敆 |
Sat 8:45 a.m. - 9:05 a.m.
|
Training physical systems with Equilibrium Propagation
(
Invited Talk
)
>
link
SlidesLive Video |
Julie Grollier 馃敆 |
Sat 9:05 a.m. - 9:25 a.m.
|
Low-precision Sampling for Probabilistic Deep Learning
(
Invited Talk
)
>
link
SlidesLive Video |
Ruqi Zhang 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Inference analysis of optical transformers ( Poster ) > link | Xianxin Guo 路 Chenchen Wang 路 Djamshid Damry 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Diffractive Optical Neural Networks with Arbitrary Spatial Coherence ( Poster ) > link | Matthew Filipovich 路 Aleksei Malyshev 路 Alexander Lvovsky 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
The Data Movement Bottleneck in Analog Computing Accelerators: An Analog Optical Fourier Transform and Convolution Accelerator Case Study ( Poster ) > link | James Meech 路 Vasileios Tsoutsouras 路 Phillip Stanley-Marbell 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Hierarchy of the echo state property in quantum reservoir computing ( Poster ) > link | Shumpei Kobayashi 路 Hoan Tran Quoc 路 Kohei Nakajima 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Contrastive power-efficient physical learning in resistor networks ( Poster ) > link | Menachem Stern 路 Sam Dillavou 路 Dinesh Jayaraman 路 Douglas Durian 路 Andrea Liu 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Energy-Based Learning Algorithms for Analog Computing: A Comparative Study ( Poster ) > link | Benjamin Scellier 路 Maxence Ernoult 路 Jack Kendall 路 Suhas Kumar 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Expanding Spiking Neural Networks With Dendrites for Deep Learning ( Poster ) > link | Mark Plagge 路 Suma Cardwell 路 Frances Chance 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Activity Sparsity Complements Weight Sparsity for Efficient RNN Inference ( Poster ) > link | Rishav Mukherji 路 Mark Schoene 路 Khaleelulla Khan Nazeer 路 Christian Mayr 路 Anand Subramoney 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Algebraic Design of Physical Computing System for Time-Series Generation ( Poster ) > link | Mizuka Komatsu 路 Takaharu Yaguchi 路 Kohei Nakajima 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
PhyFF: Physical forward forward algorithm for in-hardware training and inference ( Poster ) > link | Ali Momeni 路 Babak Rahmani 路 Matthieu Mall茅jac 路 Philipp del Hougne 路 Romain Fleury 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
A Green Granular Convolutional Neural Network with Software-FPGA Co-designed Learning ( Poster ) > link | Yanqing Zhang 路 Huaiyuan Chu 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Real-Time FJ/MAC PDE Solvers via Tensorized, Back-Propagation-Free Optical PINN Training ( Poster ) > link | Yequan Zhao 路 Xian Xiao 路 Xinling Yu 路 Ziyue Liu 路 Zhixiong Chen 路 Geza Kurczveil 路 Raymond Beausoleil 路 Zheng Zhang 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Enhancing Low-Precision Sampling via Stochastic Gradient Hamiltonian Monte Carlo ( Poster ) > link | Ziyi Wang 路 Yujie Chen 路 Ruqi Zhang 路 Qifan Song 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Squeezed Edge YOLO: Onboard Object Detection on Edge Devices ( Poster ) > link | Edward Humes 路 Mozhgan Navardi 路 Tinoosh Mohsenin 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Virtual reservoir acceleration for CPU and GPU: Case study for coupled spin-torque oscillator reservoir ( Poster ) > link | Thomas de Jong 路 Nozomi Akashi 路 Tomohiro Taniguchi 路 Hirofumi Notsu 路 Kohei Nakajima 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Adjoint Method: The Connection between Analog-based Equilibrium Propagation Architectures and Neural ODEs ( Poster ) > link | Mohamed Watfa 路 Alberto Garcia-Ortiz 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Beyond Digital: Harnessing Analog Hardware for Machine Learning ( Poster ) > link | Marvin Syed 路 Kirill Kalinin 路 Natalia Berloff 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Biologically-plausible hierarchical chunking on mixed-signal neuromorphic hardware ( Poster ) > link | Atilla Schreiber 路 Shuchen Wu 路 Chenxi Wu 路 Giacomo Indiveri 路 Eric Schulz 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Frequency propagation: Multi-mechanism learning in nonlinear physical networks ( Poster ) > link | Vidyesh Anisetti 路 Ananth Kandala 路 Benjamin Scellier 路 J. M. Schwarz 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
The Benefits of Self-Supervised Learning for Training Physical Neural Networks ( Poster ) > link | Jeremie Laydevant 路 Peter McMahon 路 Davide Venturelli 路 Paul Lott 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Towards low power cognitive load analysis using EEG signal: A neuromorphic computing approach ( Poster ) > link | Dighanchal Banerjee 路 Sounak Dey 路 Debatri Chatterjee 路 Arpan Pal 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Neuromorphic Co-Design as a Game ( Poster ) > link | Craig M Vineyard 路 William Severa 路 Brad Aimone 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Device Codesign using Reinforcement Learning and Evolutionary Optimization ( Poster ) > link | Catherine Schuman 路 Suma Cardwell 路 Karan Patel 路 J. Smith 路 Jared Arzate 路 Andrew Maicke 路 Samuel Liu 路 Jaesuk Kwon 路 Jean Anne Incorvia 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Mean-Field Assisted Deep Boltzmann Learning with Probabilistic Computers ( Poster ) > link | Shuvro Chowdhury 路 Shaila Niazi 路 Kerem Camsari 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Emergent learning in physical systems as feedback-based aging in a glassy landscape ( Poster ) > link | Vidyesh Anisetti 路 Ananth Kandala 路 Jennifer Schwarz 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Unleashing Hyperdimensional Computing with Nystr枚m Method based Data Adaptive Encoding ( Poster ) > link | Quanling Zhao 路 Anthony Thomas 路 Xiaofan Yu 路 Tajana S Rosing 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Thermodynamic AI and Thermodynamic Linear Algebra ( Poster ) > link | Patrick Coles 路 Maxwell Aifer 路 Kaelan Donatella 路 Denis Melanson 路 Max Hunter Gordon 路 Thomas Ahle 路 Daniel Simpson 路 Gavin Crooks 路 Antonio Martinez 路 Faris Sbahi 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Nonlinear Classification Without a Processor ( Poster ) > link | Sam Dillavou 路 Benjamin Beyer 路 Menachem Stern 路 Marc Miskin 路 Andrea Liu 路 Douglas Durian 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Neural Deep Operator Networks representation of Coherent Ising Machine Dynamics ( Poster ) > link | Arsalan Taassob 路 Davide Venturelli 路 Paul Lott 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Exploiting Symmetric Temporally Sparse BPTT for Efficient RNN Training ( Poster ) > link | Xi Chen 路 Chang Gao 路 Zuowen Wang 路 Longbiao Cheng 路 Sheng Zhou 路 Shih-Chii Liu 路 Tobi Delbruck 馃敆 |
Sat 9:25 a.m. - 10:30 a.m.
|
Poster Session I
(
Poster Session
)
>
|
馃敆 |
Sat 10:30 a.m. - 11:30 a.m.
|
Lunch Break
|
馃敆 |
Sat 11:30 a.m. - 11:50 a.m.
|
The Optimization Conundrum: Balancing Data, Speed, and Accuracy in Real-Time Machine Learning
(
Invited Talk
)
>
link
SlidesLive Video |
Mihaela van der Schaar 馃敆 |
Sat 11:50 a.m. - 12:00 p.m.
|
A Lagrangian Perspective on Dual Propagation
(
Oral
)
>
link
SlidesLive Video |
Rasmus H酶ier 路 Christopher Zach 馃敆 |
Sat 12:00 p.m. - 12:10 p.m.
|
Scaling of Optical Transformers
(
Oral
)
>
link
SlidesLive Video |
Maxwell Anderson 路 Shi-Yuan Ma 路 Tianyu Wang 路 Logan Wright 路 Peter McMahon 馃敆 |
Sat 12:10 p.m. - 12:30 p.m.
|
Low-Precision & Analog Computational Techniques for Sustainable & Accurate AI Inference & Training
(
Invited Talk
)
>
link
SlidesLive Video |
Kailash Gopalakrishnan 馃敆 |
Sat 12:30 p.m. - 2:00 p.m.
|
Poster Session II
(
Poster Session
)
>
|
馃敆 |
Sat 2:00 p.m. - 3:00 p.m.
|
Panel Discussion
(
Panel
)
>
SlidesLive Video |
馃敆 |