is there a notion of convergence?

[targos]

the iterations parameter litterally means that it will do X iterations. is there a way to know that the model has converged and stop before?

[jajoe]

When I read other implementations (e.g https://beckernick.github.io/neural-network-scratch/), seldom I've seen sth to stop before the end of the X iterations.

[jajoe]

Correction : if we can give a validation set to the train() function, we can monitor the convergence (by computing the accuracy on the validation set).