Workshop
ML with New Compute Paradigms
Jannes Gladrow 路 Babak Rahmani 路 Julie Grollier 路 Peter McMahon 路 Ruqi Zhang 路 Jack Kendall
West Meeting Room 114, 115
Sun 15 Dec, 9 a.m. PST
Digital computing is approaching fundamental limits and faces serious challenges in terms of scalability, performance, and sustainability. At the same time, generative AI is fuelling an explosion in compute demand. There is, thus, a growing need to explore non-traditional computing paradigms, such as (opto-)analog, neuromorphic hardware, and physical systems.Expanding on last year's successful NeurIPS workshop, which was the first of its kind in this community, we aim to bring together researchers from machine learning and alternative computation fields to establish new synergies between ML models and non-traditional hardware. Co-designing models with specialized hardware, a feature that has also been key to the synergy of digital chips like GPUs and deep learning, has the potential to offer a step change in the efficiency and sustainability of machine learning at scale. Beyond speeding up standard deep learning, new hardware may open the door for efficient inference and training of model classes that have been limited by compute resources, such as energy-based models and deep equilibrium models. So far, however, these hardware technologies have fallen short due to inherent noise, device mismatch, a limited set of compute operations, and reduced bit-depth. As a community, we need to develop new models and algorithms that can embrace and, in fact, exploit these characteristics. This workshop aims to encourage cross-disciplinary collaboration to exploit the opportunities offered by emerging AI accelerators both at training and at inference.
Schedule
Sun 9:00 a.m. - 9:20 a.m.
|
Introduction & Awards
(
Intro
)
>
SlidesLive Video |
Jannes Gladrow 馃敆 |
Sun 9:20 a.m. - 9:40 a.m.
|
Zico Kolter
(
Invited Talk
)
>
SlidesLive Video |
J. Zico Kolter 馃敆 |
Sun 9:40 a.m. - 9:50 a.m.
|
Casting hybrid digital-analog training into hierarchical energy-based learning
(
Oral
)
>
link
SlidesLive Video |
Timothy Nest 路 Maxence Ernoult 馃敆 |
Sun 9:50 a.m. - 10:00 a.m.
|
Dyadic Learning in Recurrent and Feedforward Models
(
Oral
)
>
link
SlidesLive Video |
Rasmus H酶ier 馃敆 |
Sun 10:00 a.m. - 10:20 a.m.
|
Dimitry Krotov
(
Invited Talk
)
>
SlidesLive Video |
Dmitry Krotov 馃敆 |
Sun 10:30 a.m. - 10:40 a.m.
|
Training Machine Learning Models with Ising Machines
(
Oral
)
>
link
SlidesLive Video |
Sayantan Pramanik 路 Kaumudibikash Goswami 路 Sourav Chatterjee 路 M Chandra 馃敆 |
Sun 10:40 a.m. - 11:00 a.m.
|
Azalia Misoheini
(
Invited Talk
)
>
SlidesLive Video |
Azalia Mirhoseini 馃敆 |
Sun 11:00 a.m. - 11:10 a.m.
|
Quantum Equilibrium Propagation: gradient-descent training of quantum systems
(
Oral
)
>
link
SlidesLive Video |
Benjamin Scellier 馃敆 |
Sun 11:30 a.m. - 12:00 p.m.
|
ML Panel
(
Panel
)
>
SlidesLive Video |
J. Zico Kolter 路 Dmitry Krotov 路 Azalia Mirhoseini 路 Jannes Gladrow 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Annealing Machine-assisted Learning of Graph Neural Network for Combinatorial Optimization ( Poster ) > link | Pablo Loyola 路 Kento Hasegawa 路 Andr茅s Hoyos-Idrobo 路 Kazuo Ono 路 Toyotaro Suzumura 路 Yu Hirate 路 Masanao Yamaoka 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Gaussian Process Predictions with Uncertain Inputs Enabled by Uncertainty-Tracking Microprocessors ( Poster ) > link | Janith Petangoda 路 Chatura Samarakoon 路 Phillip Stanley-Marbell 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Photonic KAN: a Kolmogorov-Arnold Network Inspired Efficient Photonic Neuromorphic Architecture ( Poster ) > link | Yiwei Peng 路 Sean Hooten 路 Thomas Van Vaerenbergh 路 Xian Xiao 路 Marco Fiorentino 路 Raymond Beausoleil 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Accelerating AI Performance using Anderson Extrapolation on GPUs ( Poster ) > link | Saleem Al Dajani 路 David Keyes 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Legendre-SNN on Loihi-2: Evaluation and Insights ( Poster ) > link | Ramashish Gaurav 路 Terrence Stewart 路 Yang Yi 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Quantum Diffusion Model for Quark and Gluon Jet Generation ( Poster ) > link | Mariia Baidachna 路 Sergei Gleyzer 路 Konstantin Matchev 路 Katia Matcheva 路 Kyoungchul Kong 路 Gopal Ramesh Dahale 路 Isabel Pedraza 路 Tom Magorsch 路 Rey Guadarrama 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Universal approximation capabilities of coherent diffractive systems ( Poster ) > link | Lennart Schlieder 路 Valentin Volchkov 路 Alexander Song 路 Peer Fischer 路 Bernhard Sch枚lkopf 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Quantum Generative Adversarial Networks for High Energy Physics Simulations ( Poster ) > link | Rey Guadarrama 路 Sergei Gleyzer 路 Konstantin Matchev 路 Katia Matcheva 路 Kyoungchul Kong 路 Gopal Ramesh Dahale 路 Mariia Baidachna 路 Haydee Hern谩ndez-Arellano 路 Isabel Pedraza 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Analog Bayesian neural networks are insensitive to the shape of the weight distribution ( Poster ) > link | Ravi G Patel 路 Tianyao Xiao 路 Sapan Agarwal 路 Christopher H Bennett 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Hardware-Algorithm Co-Design for Hyperdimensional Computing Based on Memristive System-on-Chip ( Poster ) > link | Yi Huang 路 Alireza Jaberi Rad 路 Qiangfei Xia 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Quantum Equilibrium Propagation: gradient-descent training of quantum systems ( Poster ) > link | Benjamin Scellier 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Advancing Neuromorphic Computing Algorithms and Systems with NeuroBench ( Poster ) > link | Jason Yik 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Lie-Equivariant Quantum Graph Neural Networks ( Poster ) > link | Jogi Suda Neto 路 Roy Forestano 路 Sergei Gleyzer 路 Kyoungchul Kong 路 Konstantin Matchev 路 Katia Matcheva 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
A Fully Analog Pipeline for Portfolio Optimization ( Poster ) > link | James Cummins 路 Natalia Berloff 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
SLaNC: Static LayerNorm Calibration ( Poster ) > link | Mahsa Salmani 路 Nikita Trukhanov 路 Ilya Soloveychik 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Integrated Photonic Lattice Filter for Accelerating Deep Convolutional Networks ( Poster ) > link | Matthew Filipovich 路 Folkert Horst 路 Bert Offrein 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Noise Aware Finetuning for Analog Non-Linear Dot Product Engine ( Poster ) > link | Lei Zhao 路 Luca Buonanno 路 Aishwarya Natarajan 路 Jim Ignowski 路 Giacomo Pedretti 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Event-based backpropagation on the neuromorphic platform SpiNNaker2 ( Poster ) > link | Gabriel B茅na 路 Timo Wunderlich 路 Mahmoud Akl 路 Bernhard Vogginger 路 Christian Mayr 路 Hector Gonzalez 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Improving Deep Learning Speed and Performance through Synaptic Neural Balance ( Poster ) > link | Antonios Alexos 路 ian domingo 路 Pierre Baldi 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
DQA: An Efficient Method for Deep Quantization of Deep Neural Network Activations ( Poster ) > link | Wenhao Hu 路 Paul Henderson 路 Jos茅 Cano 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
A cookbook for hardware-friendly implicit learning on static data ( Poster ) > link | Maxence Ernoult 路 Rasmus H酶ier 路 Jack Kendall 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Training Spiking Neural Networks via Augmented Direct Feedback Alignment ( Poster ) > link | Yongbo Zhang 路 Katsuma Inoue 路 Mitsumasa Nakajima 路 Toshikazu Hashimoto 路 Yasuo Kuniyoshi 路 Kohei Nakajima 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
A Diagonal State Space Model on Loihi 2 for Efficient Streaming Sequence Processing ( Poster ) > link | Svea Marie Meyer 路 Philipp Weidel 路 Philipp Plank 路 Leobardo Campos-Macias 路 Sumit Bam Shrestha 路 Philipp Stratmann 路 Mathis Richter 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Thermodynamic Bayesian Inference ( Poster ) > link | Maxwell Aifer 路 Kaelan Donatella 路 Samuel Duffield 路 Denis Melanson 路 Phoebe Klett 路 Gavin Crooks 路 Antonio Martinez 路 Patrick Coles 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
On the role of noise in factorizers for disentangling distributed representations ( Poster ) > link | Geethan Karunaratne 路 Michael Hersche 路 Abu Sebastian 路 Abbas Rahimi 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Casting hybrid digital-analog training into hierarchical energy-based learning ( Poster ) > link | Timothy Nest 路 Maxence Ernoult 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Regularizing the Infinite: Improved Generalization Performance with Deep Equilibrium Models ( Poster ) > link | Babak Rahmani 路 Jannes Gladrow 路 Kirill Kalinin 路 Heiner Kremer 路 Christos Gkantsidis 路 Hitesh Ballani 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Nanowire Neural Networks for time-series processing ( Poster ) > link | Veronica Pistolesi 路 Andrea Ceni 路 Claudio Gallicchio 路 Gianluca Milano 路 Carlo Ricciardi 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
A primer on in vitro biological neural networks ( Poster ) > link | Frithjof Gressmann 路 Ashley Chen 路 Lily Xie 路 Sarah Dowden 路 Nancy Amato 路 Lawrence Rauchwerger 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Energy-Efficient Random Number Generation Using Stochastic Magnetic Tunnel Junctions ( Poster ) > link | Nicolas Alder 路 Shivam Kajale 路 Milin Tunsiricharoengul 路 Deblina Sarkar 路 Ralf Herbrich 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
AIHWKIT-Lightning: A Scalable HW-Aware Training Toolkit for Analog In-Memory Computing ( Poster ) > link | Julian B眉chel 路 William Simon 路 Corey Lammie 路 Giovanni Acampa 路 Kaoutar El Maghraoui 路 Manuel Le Gallo 路 Abu Sebastian 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
A fast algorithm to simulate nonlinear resistive networks ( Poster ) > link | Benjamin Scellier 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Enabling On-Device Large Language Models with 3D-Stacked Memory ( Poster ) > link | Lita Yang 路 Kavya Sreedhar 路 Huichu Liu 路 Edith Beigne 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Analog Gradient Calculation of Optical Activation Function Material ( Poster ) > link | Jakub Kostial 路 Filipe Ferreira 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Deep activity propagation via weight initialization in spiking neural networks ( Poster ) > link | Aurora Micheli 路 Olaf Booij 路 Jan van Gemert 路 Nergis Tomen 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
MoQ: Mixture-of-format Activation Quantization for Communication-efficient AI Inference System ( Poster ) > link | Haonan Wang 路 Zeli Liu 路 Chao Fang 路 John Walters 路 Stephen Crago 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Improving Analog Neural Network Robustness: A Noise-Agnostic Approach with Explainable Regularizations ( Poster ) > link | Alice Duque 路 Pedro Freire 路 Egor Manuylovich 路 Sergei K. Turitsyn 路 Jaroslaw Prilepsky 路 Dmitrii Stoliarov 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Dyadic Learning in Recurrent and Feedforward Models ( Poster ) > link | Rasmus H酶ier 路 Kirill Kalinin 路 Maxence Ernoult 路 Christopher Zach 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Multi-Task Neural Network Mapping onto Analog-Digital Heterogeneous Accelerators ( Poster ) > link | Hadjer Benmeziane 路 Corey Lammie 路 Athanasios Vasilopoulos 路 Irem Boybat 路 Manuel Le Gallo 路 Hsinyu Tsai 路 Kaoutar El Maghraoui 路 Abu Sebastian 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Federated Learning with Quantum Computing and Fully Homomorphic Encryption: A Novel Computing Paradigm Shift in Privacy-Preserving ML ( Poster ) > link | Siddhant Dutta 路 Pavana Karanth 路 Pedro Maciel Xavier 路 Iago de Freitas 路 Nouhaila Innan 路 Sadok Ben Yahia 路 Muhammad Shafique 路 David Bernal Neira 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Information Bottleneck of Quantum Neural Networks ( Poster ) > link | Juexiao Wang 路 Myeongsu Kim 路 Sabre Kais 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Analog Computing for AI Sometimes Needs Correction by Digital Computing: Why and When ( Poster ) > link | Changdae Kim 路 Daegun Yoon 路 Taehoon Kim 路 Yeonjeong Jeong 路 Kangho Kim 路 Kwangwon Koh 路 Eunji Pak 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
N Multipliers for N Bits: Learning Bit Multipliers for Non-Uniform Quantization ( Poster ) > link | Raghav Singhal 路 Anmol Biswas 路 Sivakumar Elangovan 路 Shreyas Sabnis 路 Udayan Ganguly 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Training Machine Learning Models with Ising Machines ( Poster ) > link | Sayantan Pramanik 路 Kaumudibikash Goswami 路 Sourav Chatterjee 路 M Chandra 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Hyperspectral Compute-In-Memory: An Opto-Electronic Computing Architecture Enabling Compute Density Beyond PetaOPS/mm ( Poster ) > link | Myoung-Gyun Suh 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
High-speed random number generator co-processors for machine learning and AI acceleration ( Poster ) > link | Shannon Egan 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Designing Polaritonic Integrated Circuits for Quantum Processing ( Poster ) > link | Mathias Van Regemortel 路 Wolfger Peelaers 路 Thomas Van Vaerenbergh 馃敆 |
Sun 12:00 p.m. - 1:40 p.m.
|
Bulk Bitwise Accumulation in Commercial DRAM ( Poster ) > link | Tatsuya Kubo 路 Masayuki Usui 路 Tomoya Nagatani 路 Daichi Tokuda 路 Lei Qu 路 Ting Cao 路 Shinya Takamaeda-Yamazaki 馃敆 |
Sun 1:40 p.m. - 2:00 p.m.
|
Mike Davies
(
Invited Talk
)
>
SlidesLive Video |
Mike Davies 馃敆 |
Sun 2:00 p.m. - 2:10 p.m.
|
A Diagonal State Space Model on Loihi 2 for Efficient Streaming Sequence Processing
(
Oral
)
>
link
SlidesLive Video |
Svea Marie Meyer 路 Philipp Weidel 路 Philipp Plank 路 Leobardo Campos-Macias 路 Sumit Bam Shrestha 路 Philipp Stratmann 路 Mathis Richter 馃敆 |
Sun 2:10 p.m. - 2:30 p.m.
|
Clara Wanjura
(
Invited Talk
)
>
SlidesLive Video |
Clara Wanjura 馃敆 |
Sun 2:30 p.m. - 2:40 p.m.
|
Photonic KAN: a Kolmogorov-Arnold Network Inspired Efficient Photonic Neuromorphic Architecture
(
Oral
)
>
link
SlidesLive Video |
Yiwei Peng 路 Sean Hooten 路 Thomas Van Vaerenbergh 路 Xian Xiao 路 Marco Fiorentino 路 Raymond Beausoleil 馃敆 |
Sun 2:50 p.m. - 3:10 p.m.
|
Jannes Gladrow
(
Invited Talk
)
>
SlidesLive Video |
Jannes Gladrow 馃敆 |
Sun 3:10 p.m. - 3:20 p.m.
|
Bulk Bitwise Accumulation in Commercial DRAM
(
Oral
)
>
link
SlidesLive Video |
Tatsuya Kubo 路 Masayuki Usui 路 Tomoya Nagatani 路 Daichi Tokuda 路 Lei Qu 路 Ting Cao 路 Shinya Takamaeda-Yamazaki 馃敆 |
Sun 3:20 p.m. - 3:40 p.m.
|
Patrick Coles
(
Invited Talk
)
>
SlidesLive Video |
Patrick Coles 馃敆 |
Sun 4:30 p.m. - 5:00 p.m.
|
Hardware Panel
(
Panel
)
>
SlidesLive Video |
Phillip Stanley-Marbell 路 Patrick Coles 路 Jannes Gladrow 路 Mike Davies 路 Jack Kendall 路 Clara Wanjura 馃敆 |