final project of W4995 Machine Learning for functional genomics
Following show the basic usage to reproduce all experimental results. More detail usage will come soon
First, download the data and put them into a ./data/ folder
. download.sh
Then run . samlple.sh to get different percentage of labelled data in each dataset
. sample.sh
To run the first experiment, simply run
. run_ref.sh
To run the second experiment, run
. run_paired.sh
To get all tsne plots, run
python tsne_plot.py
To plot the testing accuracy figures, use acc_plot.py. Following is an example that plot the testing accuray versus epoch in 15% labelled human pancreas dataset, change the input parameter to control which experiment you want to plot and where you want to save the results.
python acc_plot.py --res ./results/baron_2016h_labelled_15/ --save_path ./
The overall time to run all stuff should be less than 2 hours on a single Titan X GPU machine