Optimal Training Subset

The goal of the project is to identify a subset of the most representative examples from each class in an image classification problem (e.g., datasets like FashionMNIST, CIFAR-10, or CIFAR-100). The objective is to determine which images are sufficient to train a well-performing classifier that ensures optimal separation between classes.

Instalation

1. Clone repository
   git clone https://gitlab-stud.elka.pw.edu.pl/mostasze/optimal_training_subset.git
2. Prepare enviroment
   make_venv
3. Install requirements
   make requirements

Running experiments

In order to replicate experiments run
make run_experiments.
To inspect results run
mlflow ui.

├── Makefile           <- Makefile with convenience commands
├── README.md          <- The top-level README for developers using this project.
├── data
│   └── raw            <- The original, immutable data dump.
│
├── docs               <- Documents
│
├── notebooks          <- Jupyter notebooks. Initial experiments.
│
├── pyproject.toml   
│
├── references         <- Data dictionaries, manuals, and all other explanatory materials.
│
├── requirements.txt   <- The requirements file for reproducing the analysis environment
│
├── setup.cfg          <- Configuration file for flake8
│
└── optimal_training_subset   <- Source code for use in this project.
    │
    ├── data           <- Data management and loading
    │
    ├── evolutionary   <- Evolutionary strategies implementations
    │
    ├── experiments    <- Experiment scripts
    │
    ├── models         <- Model definitions and architectures
    │
    ├── optimizers     <- Hill climbing algorithms
    │
    ├── utils          <- Utility functions and helpers
    │   
    └── config.py      <- Configuration settings

---

Authors

Mateusz Ostaszewski
Michał Sadowski