Equilibrium Propagation: Bridging the Gap between Energy-Based Models and Backpropagation by B Scellier and Y Bengio in Front. Comput. Neurosci. 4 May 2017

Could equilibrium propagation be nature’s way of doing backpropagation?

This paper is absolutely not for the faint-hearted. But it tackles an enormously important topic. How does one create artificial neural networks that are biologically plausible? The problem to date is that backpropagation is not biologically plausible. So, ANNs are by definition not good models of the human brain. Whilst they are immensely useful in their own right, they have serious shortcomings in the effort to create artificial human-level intelligence. The authors do not pretend that they have answered all the questions on the table. But they put forward equilibrium propagation as being a major step in, as they say, bridging the gap.

Link to paper: https://www.frontiersin.org/articles/10.3389/fncom.2017.00024/full

You might also like to browse other AI papers: https://www.thesentientrobot.com/category/ai/ai-academic-papers/