PyML

A library containing basic structure and layers for deep learning.

Note: Used code developed in Stanford CS231n for fast vectorized calculation.

Introduction

PyML, written in python, is a library that includes basic frameworks and layers for introductory level machine learning, aiming to provide easy and simple solutions to basic networks with clarity as well as comparable speed and accuracy with other popular deep learning frameworks. The layers are designed to be easily combined in order to fit different network.

File Structure:

core frame.py: the structure that integrates all layers and performs training layers: containing all layers Layer.py: superclass for all layers DenseLayer.py ReLu.py Softmax.py SVM.py ConvLayer.py MaxPool.py data data_util.py: provides easy access to all datasets contained cifar-10 mnist examples Sample files that perform deep learning using the library

Running Examples

Make sure your terminal is in directory PyML/examples, available examples: mnist_fc_net: running fully connected network on MNIST cifar10_fc_net: running fully connected network on Cifar-10 cifar10_conv_net: running convolutional neural network on Cifar-10

GUI

Make sure your terminal is in directiry PyML/GUI, type:

python user_interface.py