An implemention of MLP (Multi Layer Perceptron) Classifier from scratch in numpy.

Description

• hidden layers activation is sigmoid
• output layer activation is softmax
• loss function is cross-entropy
• support save and load model weights
• setting hidden layers by paramater "hidden_sizes", examples: hidden_sizes=(100,),there is a hidden layer of 100 neurons; hidden_sizes=(100,50),there are two hidden layers, 100 and 200 neurons.

Example:

Mnist Dataset Classfication

1. Download Data:

• mnist_train.csv
• mnist_test.csv

2. Run example

3. Training log of example