This project aims to train a multilayer perceptron (MLP) deep neural network on MNIST dataset using numpy. The MNIST dataset of handwritten digits has 784 input features (pixel values in each image) and 10 output classes representing numbers 0–9.
Download the codebase and open up a terminal in the root directory. Make sure python 3.6 is installed in the current environment. Then execute
pip install -r requirements.txt
This should install all the necessary packages for the code to run.
The file mnist_testdata.hdf5 (not uploaded here due to size constraints) contains 10,000 images in the key ‘xdata’ and their corresponding labels in the key ‘ydata’. Each image is 784-dimensional and each label is one-hot 10-dimensional. So if the label for an image is [0, 0, 0, 1, 0, 0, 0, 0, 0, 0], it means the image depicts a 3
Refer to the file Different Configurations Report.pdf for details and graphs on the various configurations of the MLP.
The best accuracy was obtained using the following configuration:
- Input Layer - 784 neurons
- 1 Hidden layer of 512 neurons
- Output Layer - 10 neurons
- Batch Size – 50
- Epoch – 50
- Learning Rate - 0.09
- Activation in Intermediate Layers – ReLu Activation
- Activation in Output Layer – Softmax
- Parameter Initialization – He Normalization
- Final Training Accuracy : 99.99%
- Final Test Accuracy : 98.26%
Models are scored on classification accuracy which is simply the number of correct classifications divided by the total number of elements in the test set.