Linfeng

The GAN loss problem

1

The GAN loss binary_cross_entropy, I think, should take logits and labels both in (0,1), Maybe the author forget to add sigmoid activation in the last dense layer of Discriminator? ?...

https://github.com/scikit-multiflow/scikit-multiflow/blob/d073a706b5006cba2584761286b7fa17e74e87be/src/skmultiflow/drift_detection/hddm_w.py#L84 https://github.com/scikit-multiflow/scikit-multiflow/blob/d073a706b5006cba2584761286b7fa17e74e87be/src/skmultiflow/drift_detection/hddm_w.py#L220 As is discribed in the title.

What does the index of alphas_network mean in auto_deeplab.py?

1

From my point of view, The index of self.alphas_network with size(12,4,3) should correspond to (layer_numbers(12), network_depth(4,8,16,32), actions(up,level,down) ). For simplicity, we can just give different index specific meaning, like for...

The network weights and the architecture weights(coefficient) train together?

2

I think it would be better if we train the network weights and the architecture weights separately, to be exact , frozen the grad of α，β when updating w, also...