Scaling Equilibrium Propagation to Deep ConvNets by Drastically Reducing its Gradient Estimator Bias
This work addresses the scalability issue of EP for hardware-efficient learning, making it a more viable alternative to backpropagation for deep neural networks.
The paper tackled the problem of scaling Equilibrium Propagation (EP) to deep convolutional networks by identifying and eliminating a bias in its gradient estimator, enabling training on visual tasks beyond MNIST with architectures including distinct forward and backward connections.
Equilibrium Propagation (EP) is a biologically-inspired counterpart of Backpropagation Through Time (BPTT) which, owing to its strong theoretical guarantees and the locality in space of its learning rule, fosters the design of energy-efficient hardware dedicated to learning. In practice, however, EP does not scale to visual tasks harder than MNIST. In this work, we show that a bias in the gradient estimate of EP, inherent in the use of finite nudging, is responsible for this phenomenon and that cancelling it allows training deep ConvNets by EP, including architectures with distinct forward and backward connections. These results highlight EP as a scalable approach to compute error gradients in deep neural networks, thereby motivating its hardware implementation.