Second Annual Data Science Bowl

Code for Second Annual Data Science Bowl. 16th place.

Sumamry

A Hybrid Deep Neural Network using CNN and MLP.

It is used only 3 SAX slices to predict the actual volume, is not used segmentation techniques, is not needed hand-labelings.

Developer Environment

• Ubuntu 14.04
• 12GB RAM
• GPU & CUDA (I used EC2 g2.2xlarge instance)
• Torch7
• Ruby
• dicom (rubygems)
• graphicsmagick (luarocks)

Installation

Install CUDA and Torch7 first. See NVIDIA CUDA Getting Started Guide for Linux and Getting started with Torch.

sudo apt-get install libgraphicsmagick-dev ruby rubygems
sudo gem install dicom
luarocks install graphicsmagick

Data

Place the data files into a subfolder ./data.

% ls ./data
test  train  train.csv  validate  validate.csv

For validation set

./run_all.sh

For test set

./run_all_test.sh

NOTICE: I used 8 g2.xlarge instances to execute this script. See comments in ./run_all_test.sh.