Convergence of energy-based learning in linear resistive networks
arXiv:2503.00349v2 Announce Type: replace Abstract: Energy-based learning algorithms are alternatives to backpropagation and are well-suited to distributed implementations in analog electronic devices. However, a rigorous...