the test should be like this

https://www.mathworks.com/matlabcentral/answers/57387-questions-about-plotperform-and-performance-graph

Your plot is typical case illustrating underfitting versus overfitting on the training data. In between, there is a moment when the performance is optimal (just before the MSE starts to increase on the validation). I don't see anything wrong with your plot (depending on the ANN architecture, optimal performance can be reached in less or more epochs). This blog post should explain the idea in more detail:

https://towardsdatascience.com/overfitting-vs-underfitting-a-complete-example-d05dd7e19765

you cannot deduce any interpretation from the MSE variation over the number of epoch.

The MSE convergence depends on

- the size of the data set,

-the size and the type of ANN that you are using to fit your application

-The stooping criterion of the algorithm (e.g. LM algorithm) that may fall in local minima which leads to an early convergence (e.g. underfitting)

Can provide more explanation for above case please

So you mean, the higher MSE value is not result of a high epoch number or vice versa.