Intuitive Explanation of Straight-Through Estimators with PyTorch Implementation

Sometimes we want to put a threshold function at the output of a layer. This can be for a variety of reasons. One of them is that we want to summarize the activations into binary values. This binarization of activations can be useful in autoencoders. However, thresholding poses a problem during backpropagation. The derivative of threshold functions is zero.