UniLTN

UniLTN - Code for the paper:

Enhancing Logical Tensor Networks: Integrating Uninorm-Based Fuzzy Operators for Complex Reasoning

Project Structure

The project structure is organized as follows:

• data: Directory for storing datasets and loading them.
• experiments: Directory for experiment configurations and results.
  • configurations: Subdirectory for experiment configuration files categorized by example (e.g., binary classification and parent_ancestor).
  • results: Subdirectory for storing experiment results categorized by example.
  • operators.py: Python module for defining operators used in experiments.
  • plots.py: Python module for generating plots to visualize experiment results.
  • get_plots_from_csv.py: Python module for generating plots from csv files of multilabel and mnist_single_digit_addition examples.
• ltn: Directory for logic tensor networks (LTN) related code.
• models: Directory containing modules for model construction and utilities.
  • axioms.py: Python module defining axioms for logic tensor networks.
  • commons.py: Python module containing train functions used across examples.
  • evaluate.py: Python module containing evaluation functions for a subset of examples.
  • models.py: Python module for defining models.
  • steps.py: Python module for defining training and test steps.
  • utils.py: Utility functions.
• main.py: Main script to execute the project.
• utils.py: General utility functions.
• uniltn.yml: YAML file specifying the conda environment required to run the project.

Running the Code

1. Set up the Project Environment

• First, ensure you have Conda installed. If not, follow the instructions here.

• Then, create the project environment using the provided fuzzyltn_env.yml file:

  conda env create -f uniltn_env.yml

• And activate it:

  conda activate uniltn

2. Set up Configuration

• Ensure you have a configuration file located in the experiments/configurations/<example_name>/ directory.

• This file should contain experiment settings such as seed, epochs and optimizer.

3. Command-line Arguments

• The main script accepts command-line arguments to specify dataset, path to configuration file, and directory to store results.

• Use the following command-line arguments:

  • -example: Specify the example to run (default is "binaryc").
  • -path_to_conf: Provide path to configuration file (default is ./experiments/configurations/<example_name>/ltn/conf-00.json).
  • -path_to_results: Define directory where results will be saved (default is ./experiments/results/<example_name>/ltn/).

4. Run the Main Script

• Run the main script:

  python ./main.py -path_to_conf ./experiments/configurations/<example_name>/ltn/conf-00.json -path_to_results ./experiments/results/<example_name>/ltn/ -example <example_name>
  

5. Get plots for Multilabel and Single Digit Addition

• Move inside experiments folder
• Run the get_plots_from_csv.py script:

  python ./get_plots_from_csv.py -example <example_name> -path_to_results <path_to_results> -metric <metric>  -exp_id <exp_id>
  

  • -example_name: name of the example to plot (i.e, multilabel, mnist_single_digit_addition, default is multilabel).
  • -path_to_results: path containing the csv files (default is multilabel).
  • -exp_id: identifier for exp (default is seed-0_epochs500_lr0.001).
  • -metric: Metric to plot (i.e., Loss, Satisfiability, Accuracy, default is Satisfiability).

Authors

• Paulo Vitor De Campos Souza: pdecampossouza@fbk.eu
• Gianluca Apriceno: apriceno@fbk.eu
• Mauro Dragoni: dragoni@fbk.eu

License

For open source projects, say how it is licensed.