CloudPred

This repository contains code for reproducing the experiments from "CloudPred: Predicting Individual Outcome From Heterogeneous Point Clouds".

Installation

First, clone this repository and enter the directory by running:

git clone https://github.com/echonet/dynamic.git
cd dynamic

CloudPred is implemented for Python 3 and can be installed by navigating to the cloned directory and running

pip install --user .

Contents

• cloudpred contains implementations of CloudPred and other methods in the manuscript
• scripts contains code to preprocess the data, run the methods, and plot results

Experiments

Simulations

The experimental results from Figure 3 can be generated by running:

scripts/run_all.sh

This scripts calls scripts/run_simulation.sh, which generates the results for Figure 1a and 1b, and scripts/run_interaction.sh, which generates the results for Figure 1c. These in turn call scripts/synthetic.py, which generates the simulated datasets.

Lupus

The data for the lupus experiments should first be obtained form the original studies. Then, to preprocess, in this directory, run:

scripts/process_lupus.py

The results for the lupus experiments are generated by running:

scripts/run_lupus.sh

Generating Figures and Results

Using the results from the previous scripts, the performance plots in the manuscript can be generated by running:

scripts/plot.py

and the clustering results can be generated by running:

scripts/visualize.py