mlsanityData to 7 Oct 2026

Scaling of hardware-compatible perturbative training algorithms

arXiv 2501.15403 ↗First posted 26 Jan 2025cs.LG cs.NE math.OCcited by 0

Not on the map. Why?

Abstract (arXiv)

In this work, we explore the capabilities of multiplexed gradient descent (MGD), a scalable and efficient perturbative zeroth-order training method for estimating the gradient of a loss function in hardware and training it via stochastic gradient descent. We extend the framework to include both weight and node perturbation, and discuss the advantages and disadvantages of each approach. We investigate the time to train networks using MGD as a function of network size and task complexity. Previous research has suggested that perturbative training methods do not scale well to large problems, since in these methods the time to estimate the gradient scales linearly with the number of network parameters. However, in this work we show that the time to reach a target accuracy--that is, actually solve the problem of interest--does not follow this undesirable linear scaling, and in fact often decreases with network size. Furthermore, we demonstrate that MGD can be used to calculate a drop-in replacement for the gradient in stochastic gradient descent, and therefore optimization accelerators such as momentum can be used alongside MGD, ensuring compatibility with existing machine learning practices. Our results indicate that MGD can efficiently train large networks on hardware, achieving accuracy comparable to backpropagation, thus presenting a practical solution for future neuromorphic computing systems.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 18 of its 55 references that are in the map. First posted on arXiv 26 Jan 2025.

Builds on

18 of 55 references.

References · 18

  1. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  2. 2023Fine-Tuning Language Models with Just Forward Passescited by 337
  3. 2016Equilibrium Propagation: Bridging the Gap Between Energy-Based Models and Backpropagationcited by 251
  4. 2016Direct Feedback Alignment Provides Learning in Deep Neural Networkscited by 240
  5. 2021Deep physical neural networks enabled by a backpropagation algorithm for arbitrary physical systemscited by 125
  6. 2018Assessing the Scalability of Biologically-Motivated Deep Learning Algorithms and Architecturescited by 100
  7. 2023DeepZero: Scaling up Zeroth-Order Optimization for Deep Model Trainingcited by 76
  8. 2022Scaling Forward Gradient With Local Lossescited by 65
  9. 2018Eligibility Traces and Plasticity on Behavioral Time Scales: Experimental Support of neoHebbian Three-Factor Learning Rulescited by 64
  10. 2024Training of Physical Neural Networkscited by 47
  11. 2022Experimentally realized in situ backpropagation for deep learning in nanophotonic neural networkscited by 43
  12. 2023Hardware-aware training for large-scale and diverse deep learning inference workloads using in-memory computing-based acceleratorscited by 42
6 more
  1. 2023Brain-inspired learning in artificial neural networks: a reviewcited by 26
  2. 2023Machine Learning Without a Processor: Emergent Learning in a Nonlinear Electronic Metamaterialcited by 19
  3. 2023Tensor-Compressed Back-Propagation-Free Training for (Physics-Informed) Neural Networkscited by 12
  4. 2023Multiplexed gradient descent: Fast online training of modern datasets on hardware neural networks without backpropagationcited by 8
  5. 2016Demonstrating Hybrid Learning in a Flexible Neuromorphic Hardware Systemcited by 7
  6. 2024Online Pseudo-Zeroth-Order Training of Neuromorphic Spiking Neural Networkscited by 2

Built on by

Not cited yet.