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