Fetching the paper…

Safe Mutations for Deep and Recurrent Neural Networks through Output Gradients · Around