cs.LGJul 15, 2024

Correlations Are Ruining Your Gradient Descent

Authors: Joshua Offergeld, Nasir Ahmad

Organizations: Department of Machine Learning and Neural Computing, Donders Institute for Brain, Cognition and Behaviour, Thomas van Aquinostraat 4, Nijmegen, 6525GD, The Netherlands

Abstract

Natural gradient descent, and the preconditioned optimizers which have followed from it, correct gradient descent by estimating curvature and applying its inverse to the gradient. For any single layer of a deep neural network, this correction can be shown to approximately separate into two factors: one which accounts for correlations in the gradient signal, and one which accounts for correlations in the data arriving at that layer. Herein we show that this second factor can be expressed as a change of basis rather than of curvature and, considered thus, it can be removed from the optimizer entirely and placed within the network itself. Specifically, we show that, once forward weights are corrected, right-sided natural gradient descent can be implemented in a form which has three exactly equivalent forms: a gradient preconditioner, a decorrelating linear transform in the forward pass, or a set of recurrent dynamics in which all updates are Hebbian and rely only upon locally available signals. The (recurrent) decorrelation method we discover aligns closely with proposed methods for inhibitory learning and homeostatic plasticity in the brain. Notably, our transform requires no matrix inverse, no matrix decomposition, and no information beyond that which is available at an individual synapse. It can therefore be applied in distributed and physically constrained systems, including neuromorphic hardware and nervous systems, where existing preconditioners cannot. We demonstrate its effectiveness empirically in convolutional and residual network models while significantly improving a range of approximate biologically plausible methods for credit assignment.

Figures & tables

Appendix figures & tables7 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. Sven: Singular Value Descent as a Computationally Efficient Natural Gradient Method

    Apr 1, 2026Samuel Bright-Thonney, Thomas R. Harvey, Andre Lukas +1Deep Learning OptimizationNatural Gradient Methods

  2. Natural gradient descent with momentum

    Apr 16, 2026Anthony Nouy, Agustín SomacalDeep Learning OptimizationRiemannian Optimization

  3. Natural Gradient Learning in Predictive Coding through Lateral Circuits

    Nov 12, 2026Lhea Beumer, Sander de Haan, Niclas Pokel +2Predictive CodingLocal Learning