Optimal Samples Selection System

A Python desktop application that solves the Optimal Samples Selection combinatorial optimization problem via Integer Linear Programming (ILP), with a PyQt5 GUI and SQLite result storage.

Documentation

• User Manual: docs/User_Manual.md
• Project Report: docs/Project_Report.md

Quick Start

1. (Recommended) Create and activate a virtual environment.

2. Install dependencies:

   pip3 install -r requirements.txt

3. Run the GUI:

   python3 main.py

Alternatively:

./run.sh

Project Structure

• core/: optimization solver (ILP / OR-Tools)
• gui/: PyQt5 desktop UI
• database/: SQLite persistence
• results/: saved .db result files
• docs/: documentation

Known Cover Cache

The application uses results/known_covers.sqlite as a reusable cache for:

• La Jolla Covering Repository standard covers C(v,k,t) when s == j.
• Previously solved project instances (n,k,j,s) mapped onto canonical samples 1..n.

To refresh the La Jolla cache for the assignment parameter range:

python3 scripts/import_ljcr_covers.py --min-v 7 --max-v 25 --min-k 4 --max-k 7 --min-t 3 --max-t 7

Cached La Jolla results are returned directly only when they are proven optimal. Otherwise they are used as feasible upper bounds / solver hints.

Large Instances

The GUI estimates the optimized coverage-entry count before starting an exact local solve. Instead of checking every j-subset against every k-group, the solver generates only the j-subsets each k-group can actually cover. If the estimate is too large, it avoids constructing the full model to prevent freezes or crashes. In that case, it returns a cached cover when available, or asks the user to reduce the parameters.

Mobile (Android/iOS)

• Source: mobile/
• Notes: offline exact solving (Branch-and-Bound) can be slow for larger n.
• Build instructions: mobile/README.md

Windows Installer

• Packaging scripts: packaging/windows/
• CI workflow: .github/workflows/windows-installer.yml (PyInstaller + Inno Setup)