ECE901Project

A repository containing code and reports for UW-Madison ECE901: Large-Scale Machine Learning

Setup

Make sure you are running Python 3.5.

Also, run the following commands in conda environment to update TF to the latest version.

(tensorflow) $ export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-0.12.0rc0-py3-none-any.whl
(tensorflow) $ pip install --ignore-installed --upgrade $TF_BINARY_URL

Structure

Hardware

• VGGNet-16: Contains the Vivado project for FPGA implementation.

LaTeXStyleFiles: Contains LaTeX style files that may need to be installed on your system to compile LaTeX source.

PaperPresentation: Contains LaTeX source for paper presentation assignment.

ProjectProposal: Contains LaTeX source for project proposal assignment.

ProposalPresentation: Contains LaTeX source for project proposal presentation.

Tensorflow

• TFMechanics101Tutorial: Contains source code for TF tutorial (https://www.tensorflow.org/versions/master/tutorials/mnist/tf/index.html)
• fully_connected_feed.py: Run this using python fully_connected_feed.py to train the network.
• input_data.py: Just for reference. The training code pulls this file in via import.
• mnist.py: Just for reference. The training code pulls this file in via import.
• TFCNNTutorial: Contains source code for TF CNN tutorial (https://www.tensorflow.org/versions/master/tutorials/deep_cnn/index.html)
• cifar10.py: Just for reference. The training code pulls this file in via import.
• cifar10_input.py: Just for reference. The model code pulls this file in via import.
• cifar10_train.py: Run this using python cifar10_train.py to train the network.
• TwoLayerCNN: Contains source code for CPU implementation of custom two layer CNN.
• model.py: Contains the model related functions like inference(), loss(), and train().