An algorithm to compute the softmax layer in neural networks using low floating-point precision arithmetic.
See the pdf file for the algorithm.
An algorithm to compute the softmax layer in neural networks using low floating-point precision arithmetic.