Fall 2018 Machine Learning Final Project

Team 2 - Caio, Kenneth, Michael

Quick Start

1. Clone this GitHUB repository

git clone git@github.com:taoyilee/ml_final_project.git

2. Create Python virtual environment

virtualenv venv

3. Activate Python virtual environment

source venv/bin/activate

4. Install requirement

pip install -r requirements.txt

5. Download The Street View House Numbers (SVHN) dataset
6. Traing an autoencoder with train_autoencoder.py

Special instruction for windows

To setup GPU support flow, following software packages need to be installed on your Windows operating system

• NVIDIA® GPU drivers —CUDA 9.0 requires 384.x or higher.
• CUDA® Toolkit —TensorFlow supports CUDA 9.0.; Note that CUDA includes CUPTI.
• cuDNN SDK (>= 7.2)

To make sure tensoflow is able to access all DLLs from CUDA, confirm the following environment variables are added to your PATH environment variable

SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0\bin;%PATH%
SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.0\extras\CUPTI\libx64;%PATH%
SET PATH=C:\tools\cuda\bin;%PATH%

Configuration File

Sample configuration file is shown below

[general]
train_ratio = 98
batch_size = 512
;available options: cnn, mlp
ae_model = cnn
;ae_model = mlp

;early_stop = False
early_stop = True
;color_mode grayscale, rgb
;color_mode = rgb
color_mode = grayscale

epoch = 30

noise_ratio = 0.07
dev_set = dataset_split/arrays/dev/rgb_all.npy
training_set = dataset_split/arrays/training/rgb_all.npy
dev_set_csv = dataset_split/dev.csv
training_set_csv = dataset_split/trn.csv

[plot]
tag = 12_02_220110_grayscale_mlp

[training]
patientce_reduce_lr = 4
verbosity = 1

[optimizer]
lr = 0.01

[ae_cnn]
regularization = 1e-10
filter_number = 16
hidden_layers = 3
bottleneck_filter_number = 4
; filter_size = 3 means 3x3
filter_size = 3

[ae_mlp]
bottleneck_width = 2048
hidden_layers = 0
expand_ratio = 0.5
;regularization = 1

Dataset

SVHN dataset is available at http://ufldl.stanford.edu/housenumbers/. The authors acknowledge the contributors of SVHN dataset. Below is the original publication from Andrew Y. Ng's research team at Stanford.

Yuval Netzer, Tao Wang, Adam Coates, Alessandro Bissacco, Bo Wu, Andrew Y. Ng Reading Digits in Natural Images with Unsupervised Feature Learning NIPS Workshop on Deep Learning and Unsupervised Feature Learning 2011. (PDF)

LICENSE

Copyright 2018 Tao-Yi Lee, Caio Batista de Melo, Kenneth Stewart

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.