Obtaining abnormal changes in loss and accuracy

vision
1

I am training a multi-class classification problem.Instead of declining at a steady rate, y loss and accuracy metrics are moving in zig-zag rate. I am not able to figure out the reason behind this behaviour. Can anyone please help me out?

image
image850×602 26.7 KB

If I keep on running my model for longer epochs, then the same zigzag behaviour is being obtained.

2

Try lowering your learning rate .
It will be helpful if you can upload some code snippets

1 Like
3

I did , from 0.1 to 0.0001.
My code is here - https://gist.github.com/droid786/4069fe9cb16e909eb09ed087b427fa26