microsoft / CNTK

Microsoft Cognitive Toolkit (CNTK), an open source deep-learning toolkit
https://docs.microsoft.com/cognitive-toolkit/
Other
17.49k stars 4.3k forks source link

An increase in loss after shuffling the training set #3792

Closed faruknane closed 4 years ago

faruknane commented 4 years ago

My training data -> x_train has 32 instances (32 training example).

I train my model with x_train and loss goes to 0 as below. image

After I shuffle my x_train. image

Loss goes so high. image

Why is that? I made sure that shuffling works just fine and I don't understand why this is happening.

My model is like below: image