← Search

Javier Sagastuy-Brena

2 accepted papers

2021

Neural Mechanics: Symmetry and Broken Conservation Laws in Deep Learning Dynamics

ICLR 2021poster

Understanding the dynamics of neural network parameters during training is one of the key challenges in building a theoretical foundation for deep learning. A central obstacle is that the motion of a network in high-dimensional parameter space undergoes discrete finite steps along complex stochastic…

2020

Two Routes to Scalable Credit Assignment without Weight Symmetry

ICML 2020poster

The neural plausibility of backpropagation has long been disputed, primarily for its use of non-local weight transport — the biologically dubious requirement that one neuron instantaneously measure the synaptic weights of another. Until recently, attempts to create local learning rules that avoid we…