GP4C

This folder contains the code for the paper Variational Inference for Gaussian Processes with Panel Count Data

Dependency:

1. GPML Matlab Code version 4.1, http://www.gaussianprocess.org/gpml/code/matlab/doc/
2. Limited-memory projected quasi-Newton algorithm, http://proceedings.mlr.press/v5/schmidt09a/schmidt09a.pdf
3. Safe computation of logarithm-determinat of large matrix by Dahua Lin, MIT.

The files in the main folder are main functions to run all the experiments.

1. test_synthetic_fromGP.m Use a GP to draw an intensity function and use it as the underlying intensity function to test GP4C(b=0,1,0.3).

2. test_synthetic_stepfunction.m Use a step function as the underlying intensity function to test GP4C(b=0,1,0.3).

3. test_synthetic_pseudoinputs.m Use a step function as the underlying intensity function and vary the number of pseudo inputs.

4. test_synthetic_fileration.m Use a step function as the underlying intensity function and vary the number of training files.

5. test_synthetic_duplicate.m Change the number of duplicate points in the training files and compare the computational time.

6. test_realworld_ver0.m Test GP4C,GP4CW and LocalEM on the three real-world data sets.

7. test_realworld_demo.m A demo for the realworld dataset.

8. test_additional_KS_GP3 Compare kernel smoothing and GP3.

The folder in util_plot contains the plotting functions for all the figures in the paper.