This repository contains the code used to generate the results described in the paper Application of Deep Learning to the Random Walk Theory by Matt Liston

1. install Unbuntu 14.04 on a machine with a GPU
2. install cuda version 7.5, run "sudo nvidia-modprobe -c 0 -u" to generate /dev/nvidia-uvm
3. install docker
4. install git
5. git clone https://github.com/mattliston/randomwalk.git
6. cd randomwalk
7. docker build -t ml .
8. ./launch_ml
9. python fetch.py --start 2016-06-02 --end 2006-06-02
10. h5ls nyse_nasdaq.hdf5
11. python window.py --window 100 --test_split 0.2
12. h5ls train*.hdf5 test*.hdf5
13. ls train.flist test.flist
14. caffe train -solver=solver.prototxt |& tee train.log
15. /tmp/caffe/tools/extra/parse_log.py train.log .
16. python cluster.py --window 100 --test_split 0.2 --k 50 --minit points
17. ls cluster_daily.csv cluster_summary.csv