In this notebook there are:
- Digit MNIST dataset
- Tools to visualize the samples
- Model: fully connected neural network with two hidden layer
- A train/validation function
- A random search for the hyperparameters (best model selection)
- Testing
Note: for our objectives the focus is on the model and on the function of the train. The testing part is important for us only for the output graph (what we need to get). Our dataset of interest is only the one related to the train and not validation or testing datasets: our goal is to minimize the loss function not to see the generalization of the model. So, in the following code there are some parts that we should NOT implement.