caffe training loss does not converge

Question

I'm getting the problem of non-converged training loss. (batchsize: 16, average loss:10). I have tried with the following methods + Vary the learning rate lr (initial lr = 0.002 cause very high loss, around e+10). Then with lr = e-6, the loss seem to small but do not converge. + Add initialization for bias + Add regularization for bias and weight

This is the network structure and the training loss log

Hope to hear from you Best regards

Antoine

try learning rate between 0 002 and e-6 – Caressa

I have tried with different learning rate values such as 2e-3, 2e-4,2e-5. Those values caused very high loss (e+21). – Antoine

You need to debug your training process. set debug_info: true in your 'solver.prototxt' and use these guidelines to see what is interfering with the training. – Caressa

Thank Shai for the suggestion – Antoine