# Analysis of Training Loss and Validation Loss Graph

Here I am Showing Two Loss graphs of an Artificial Neural Network.

Model 1

Model 2

Blue -training loss

Red -val training loss

Can you help me to analyse these graphs? I read some articles and post but doesn't give me any sense.

• The second graph definitely looks better..I suggest you go over to Andrew ngs course "Hyperparameter optimisation", you'll get it on Coursera or YouTube. – user9947 Jun 21 '19 at 10:34

Simply model 1 is a better fit compared to model 2.

• Graph for model 1

We notice that the training loss and validation loss aren't correlated. This means the as the training loss is decreasing, the validation loss remains the same of increases over the iterations. This means that the model is not exactly improving, but is instead overfitting the training data. This isn't what we are looking for.

• Graph for model 2

In this case, there is clearly a health correlation between training loss and the validation loss. They both seem to reduce and stay at a constant value. This means that the model is well trained and is equally good on the training data as well as the hidden data.

You should stick with model 2. In case you're going ahead with model 1, make sure to use the chechpoint where both the losses are at a similar value (at around 100 -150 epochs)

• In graph 2, is there any chance of overfitting? – Majo_Jose Jun 24 '19 at 5:26