Non-convex Learning via Replica Exchange Stochastic Gradient MCMC

Open paper
Benchmark Model Rank Results
image-classification-on-cifar-100WRN-28-10 with reSGHMC#103Percentage correct: 84.38
image-classification-on-cifar-100WRN-16-8 with reSGHMC#122Percentage correct: 82.95
image-classification-on-cifar-100ResNet56 with reSGHMC#162Percentage correct: 80.14
image-classification-on-cifar-100ResNet32 with reSGHMC#182Percentage correct: 76.55
image-classification-on-cifar-100ResNet20 with reSGHMC#191Percentage correct: 74.14
image-classification-on-cifar-10WRN-28-10 with reSGHMC#106Percentage correct: 97.42
image-classification-on-cifar-10WRN-16-8 with reSGHMC#119Percentage correct: 96.87
image-classification-on-cifar-10ResNet56 with reSGHMC#142Percentage correct: 96.12
image-classification-on-cifar-10ResNet32 with reSGHMC#161Percentage correct: 95.35
image-classification-on-cifar-10ResNet20 with reSGHMC#179Percentage correct: 94.62

Rank counts only results with a code link.