When training an RBM, we used the above to figure out how to update the weights (contrastive divergence).
We are now being asked to update the bias and not update the weights to train. The hint was to rework this partial derivative but with respect to the bias rather than the weights. Could anyone provide some help on this? Thanks.