This repository is the implementation of GPML.

The dataset should be placed inside a directory (e.g., GAMMA/) with the following structure:

GAMMA/
├── fundus_train.npy
├── fundus_test.npy
├── oct_train.npy
├── oct_test.npy
├── labels_train.npy
└── labels_test.npy

• fundus_train.npy, fundus_test.npy — Fundus image modality (training and test sets)
• oct_train.npy, oct_test.npy — OCT image modality (training and test sets)
• labels_train.npy, labels_test.npy — Classification labels (primary)

To run the main training script:

python Main.py --r $randomseed

Replace $randomseed with an integer seed (e.g., 1, 42).

Experiment settings are defined in the config JSON file.

{
  "project": "Test",
  "dataDir": "GAMMA",
  "Class": 2,
  "r": 3,
  "population_size": 100,
  "generation": 50,
  "cxProb": 0.8,
  "mutProb": 0.19,
  "elitismProb": 0.01,
  "initialMinDepth": 2,
  "initialMaxDepth": 6,
  "maxDepth": 8,
  "tournament_size": 5,
  "Strategy": "voting",
  "K": 5,
  "des": ""
}

You can change parameters such as:

• dataDir — Path to the dataset folder
• Class — Number of classes
• population_size — GP population size
• generation — Number of generations to evolve
• cxProb, mutProb — Crossover and mutation probabilities
• elitismProb — Elitism probability
• maxDepth — Maximum tree depth

This repository is built on reusing codes of https://github.com/YingBi92/BookCode