MLP Classifier

A Handwritten Multilayer Perceptron Classifier

This python implementation is an extension of artifical neural network discussed in Python Machine Learning and Neural networks and Deep learning by extending the ANN to deep neural network & including softmax layers, along with log-likelihood loss function and L1 and L2 regularization techniques.

Some Basics

An artificial neuron is mathematical function conceived as a model of biological neurons. Each of the nodes in the diagram is a a neuron, which transfer their information to the next layer through transfer function.

The transfer function is a linear combination of the input neurons and a fixed value - bias (threshold in figure). The coefficients of the input neurons are weights.
In the code, bias is a numpy array of size(layers-1) as input layer do not have a bias. The weights, also a numpy array, form a matrix for every two layers in the network.
Activation function is the output of the given neuron.

X: vectorize{(j-1)th layer}

    w = weights[j-1]
    bias = threshold[j-1]
    transfer_function = dot_product(w, X)
    o = activation(transfer_function + bias)

Details

The implementation includes two types of artificial neurons:

• Sigmoid Neurons
• Softmax Neurons

The loss function associated with Softmax function is the log-likelihood function, while the loss function for Sigmoid function is the the cross-entropy function. The calculus for both loss functions have been discussed within the code.

Further, the two most common regularization techiques - L1 and L2 have been used to prevent overfitting of training data.

Why Softmax?

For z^L_j in some vector Z^L, softmax(z^L_j) is defined as

The output from the softmax layer can be thought of as a probability distribution.

In many problems it is convenient to be able to interpret the output activation O(j) as the network's estimate of the probability that the correct output is j.

Refer these notes for calculus of softmax function.
Source of MNIST training data-set.