Overview

This project trains 3 layer feedforward neural network over a dataset of MovieLens-based user ratings to generate movie recommendations. This the code for 'Build a Movie Recommender' on Youtube.

Dependencies

[Just run this on the preconfigured AWS image so you don't have to worry about this -- these dependencies aren't trivial]

• GCC
• CuBLAS
• Cuda Toolkit (>= 7.0 is required)
• OpenMPI
• NetCDF
• JsonCPP
• CUB

Basic Usage

1. Create a GPU instance in the US East N. Virginia region using the preconfigured AMI [Amazon Machine Image] called 'ami-d6f2e6bc'
2. Download this repo, then upload it to the root directory of that instance via FileZilla
3. SSH into that instance and compile this project via the following commands

cd amazon-dsstne/src/amazon/dsstne
#Add the mpiCC and nvcc compiler in the path
export PATH=/usr/local/openmpi/bin:/usr/local/cuda/bin:$PATH
make
export PATH=`pwd`/bin:$PATH
cd samples
g++ commands.cpp demo.cpp 
./a.out

Demo.cpp is the code from the video tutorial. It'll generate a set of recommendations from the dataset in the recs file.

Credits

Credit for this library goes to the DSSTNE team at Amazon. I've merely created a C++ wrapper around all of the important functions to get people started.