NN-from-scratch

A neural network implemented using only Numpy library. Built for learning purposes.

---

The neural network lets the user get in touch with the real structure of a basic neural network, without using any deep learning library. With the parameters is it possible to control the behavior of the data and the network itself to explore different combinations and see concretely how it fits a simple batch of data.

Here a list of the settable parameters:

• learning rate --lr

• number of epochs --epochs

• complexity of data --complexity (values among 1, 2, 3, 4 or 5)

• number of points to fit --points

• seed for pseudo random data generation --seed

• number of hidden layers --hidden

• number of neurons per each hidden layer --neurons

• activation function --activation (values among 'Tanh', 'ReLU', 'Sigmoid')