MATILDA

MATILDA is a logical rule discovery system from relational databases.

🚀 Quick Start with CLI

MATILDA now features a unified command-line interface for all operations:

# General help
python cli.py --help

# Main commands
python cli.py validate --auto                    # Validate metrics
python cli.py benchmark --algorithm spider       # Run benchmark
python cli.py metrics --all                      # Calculate metrics
python cli.py test --all                         # Run tests
python cli.py clean --cache                      # Clean up
python cli.py report --latex                     # Generate reports
python cli.py info --scripts --results           # Project information

---

📦 Installation

Installation

Download database requirements

macOS:

brew install mysql-client@8.4

Linux:

sudo apt-get install mysql-client-8.4

Windows:

choco install mysql --version=8.4

Set up Python3 virtual environment

python3 -m venv venv
source venv/bin/activate  # On Windows use `venv\Scripts\activate`
pip3 install -r requirements.txt

Run the application

1. Activate the virtual environment:

source venv/bin/activate  # On Windows use `venv\Scripts\activate`

2. Start the application:

cd src 
python main.py

Optional: Run AnyBURL

To use the AnyBURL algorithm, download the AnyBURL jar from the official page and either:

• place it under src/algorithms/bins/anyburl/ (any *.jar filename), or
• set the environment variable ANYBURL_JAR to the absolute path of the jar.

Then set algorithm.name: "ANYBURL" in your config and run as usual.

---

📊 Benchmarking & LaTeX Tables (NEW!)

Need to benchmark MATILDA for a publication?

One-Click Solution ⭐

# Complete benchmark: all algorithms × all datasets × N runs + LaTeX table
python run_full_benchmark.py --runs 5

What it does:

• ✅ Runs MATILDA, SPIDER, ANYBURL, POPPER on all datasets
• ✅ Repeats N times, calculates mean ± standard deviation
• ✅ Generates professional LaTeX table automatically
• ✅ Saves results and statistics as JSON

Duration: 1-2 hours (5 runs)
Output: data/output/benchmark_table_*.tex

Three Available Scripts

Script	Usage	Speed	Statistics
run_full_benchmark.py	Complete benchmark (all algos)	🐢🐢 1-4h	✅ Yes
run_benchmark.py	Benchmark 1 algo with stats	🐢 5-30 min	✅ Yes
generate_latex_table.py	Table from existing results	⚡ < 1s	❌ No

Examples

# Complete benchmark for publication (recommended)
python run_full_benchmark.py --runs 5

# Quick benchmark (test)
python run_full_benchmark.py --runs 3 --algorithms MATILDA SPIDER

# Instant table from existing results
python generate_latex_table.py --detailed

# With configuration file
python run_full_benchmark.py --config benchmark_config.yaml

Complete Documentation

• benchmark_config.yaml - Configuration example

---