Convergence Results for Neural Networks via Electrodynamics
conference on innovations in theoretical computer science, pp. 22:1-22:19, 2018.
We study whether a two neural network can learn another depth two network using gradient descent. Assuming a linear output node, we show that the question of whether gradient descent converges to the target function is equivalent to the following question in electrodynamics: Given k fixed protons in R^d, and k electrons, each moving due ...More
Full Text (Upload PDF)
PPT (Upload PPT)