1 paper · 1 filter
Aaron Pache, Mark CW van Rossum
When training neural networks for classification tasks with backpropagation, parameters are updated on every trial, even if the sample is classified correctly. In contrast, humans…