arXiv cs.LG
· Papers
Training Non-Differentiable Networks via Optimal Transport
arXiv:2605.01928v2 Announce Type: replace Abstract: We optimize losses that jump: spiking thresholds, quantized layers, and discrete routing put jumps in the forward pass, where backpropagation does not apply. Finite differences fail: at a derivative-estimating radius, 99.5% of probe pairs on a quantized network leave