mlsanityData to 7 Oct 2026

Agnostic Physics-Driven Deep Learning

arXiv 2205.15021 ↗First posted 30 May 2022cs.LGcited by 14

Not on the map. Why?

Abstract (arXiv)

This work establishes that a physical system can perform statistical learning without gradient computations, via an Agnostic Equilibrium Propagation (Aeqprop) procedure that combines energy minimization, homeostatic control, and nudging towards the correct response. In Aeqprop, the specifics of the system do not have to be known: the procedure is based only on external manipulations, and produces a stochastic gradient descent without explicit gradient computations. Thanks to nudging, the system performs a true, order-one gradient step for each training sample, in contrast with order-zero methods like reinforcement or evolutionary strategies, which rely on trial and error. This procedure considerably widens the range of potential hardware for statistical learning to any system with enough controllable parameters, even if the details of the system are poorly known. Aeqprop also establishes that in natural (bio)physical systems, genuine gradient-based statistical learning may result from generic, relatively simple mechanisms, without backpropagation and its requirement for analytic knowledge of partial derivatives.

Key facts. Cited by 14 papers in the map (data through 7 Oct 2026). Builds on 14 of its 33 references that are in the map. First posted on arXiv 30 May 2022.

Builds on

14 of 33 references.

References · 14

  1. 2017Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithmscited by 4,283
  2. 2015BinaryConnect: Training Deep Neural Networks with binary weights during propagationscited by 571
  3. 2016Equilibrium Propagation: Bridging the Gap Between Energy-Based Models and Backpropagationcited by 251
  4. 2021Deep physical neural networks enabled by a backpropagation algorithm for arbitrary physical systemscited by 125
  5. 2020Scaling Equilibrium Propagation to Deep ConvNets by Drastically Reducing its Gradient Estimator Biascited by 60
  6. 2020Training End-to-End Analog Neural Networks with Equilibrium Propagationcited by 57
  7. 2019Updates of Equilibrium Prop Match Gradients of Backprop Through Time in an RNN with Static Inputcited by 32
  8. 2020EqSpike: Spike-driven Equilibrium Propagation for Neuromorphic Implementationscited by 29
  9. 2020Equilibrium Propagation with Continual Weight Updatescited by 25
  10. 2021A deep learning theory for neural networks grounded in physicscited by 23
  11. 2021A contrastive rule for meta-learningcited by 14
  12. 2022Learning by non-interfering feedback chemical signaling in physical networkscited by 10
2 more
  1. 2016Towards deep learning with spiking neurons in energy based models with contrastive Hebbian plasticitycited by 4
  2. 2017Mitigating Asymmetric Nonlinear Weight Update Effects in Hardware Neural Network based on Analog Resistive Synapsecited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 1; 2023: 2; 2024: 5; 2025: 3; 2026: 3.

14 papers.

Sort:

  1. 2024Training of Physical Neural Networkscited by 47
  2. 2023Energy-based learning algorithms for analog computing: a comparative studycited by 36
  3. 2022Frequency propagation: Multi-mechanism learning in nonlinear physical networkscited by 11
  4. 2023Physics-aware Machine Learning Revolutionizes Scientific Paradigm for Machine Learning and Process-based Hydrologycited by 11
  5. 2024Quantum Equilibrium Propagation: Gradient-Descent Training of Quantum Systemscited by 9
  6. 2024Quantum Equilibrium Propagation for efficient training of quantum systems based on Onsager reciprocitycited by 9
  7. 2025Towards the Training of Deeper Predictive Coding Neural Networkscited by 7
  8. 2025Learning long range dependencies through time reversal symmetry breakingcited by 6
  9. 2024A Fast Algorithm to Simulate Nonlinear Resistive Networkscited by 5
  10. 2024Towards training digitally-tied analog blocks via hybrid gradient computationcited by 3
  11. 2026Training a Predictive Coding Network on ImageNet using Equilibrium Propagationcited by 1
  12. 2025Energy-Efficient Supervised Learning with a Binary Stochastic Forward-Forward Algorithmcited by 0
2 more
  1. 2026Equilibrium Propagation for Non-Conservative Systemscited by 0
  2. 2026Optical Implementation of Equilibrium Propagation Using Spatial Photonic Ising Machinescited by 0