During Model training I am getting fluctuating accuracy in both the training accuracy and the validation accuracy .Can anyone please tell me why am i getting this and how can I fix this
Hi, @SHAHID_KAMAL !
The reason behind this behaviour is that the gradient is driving the optimization algorithm to a worse point than before (check that the loss is also increasing). One common way to avoid this is reducing the learning rate, either since the beginning or gradually during training.