Recomender 🚀

This repository houses the recomender Python package, a playful twist on "recommender". The essence of recomender is to enhance (mend) recommendation processes using Reinforcement Learning (RL) techniques.

Architectural Concept

graph TD;
    A(Model Creation Module) -->|1. Define Model A with Keras, Torch, etc.| B(RLHF Processing Module)
    B -->|2. Apply RLHF| C(Reward Generator Module)
    C -->|3. Generate Rewards| B
    B -->|4. Interact with Gym Environment| D(Gym Environment)
    A --> E(Keras, Torch for Model Definition and Training)
    C -->|5. Use Clustering for Reward Calculation| F(Clustering Method)
    D -.->|6. Provide Simulated Feedback| B
    E -.->|7. Define & Train Model A| A
    F -.->|8. Calculate Distance-Based Rewards| C

    classDef library fill:#f9f,stroke:#333,stroke-width:4px;
    classDef external fill:#bbf,stroke:#333,stroke-width:4px;
    class A,B,C library;
    class E,F,D external;

Loading

Package Architecture

/
├── examples/
│   ├── distance_based_reward.ipynb
│   └── sklearn_nn_model.ipynb
├── recomender/
│   ├── __init__.py
│   ├── builders/
│   │   ├── __init__.py
│   │   ├── DistanceBasedRewardBuilder.py
│   │   └── SklearnNNModelBuilder.py
│   └── interfaces/
│       ├── __init__.py
│       ├── RewardBuilderAbstract.py
│       └── SLModelBuilderAbstract.py
└── tests/
    └── builders/
        ├── __init__.py
        ├── test_reward_builder.py
        └── test_sklearn_nn_model_builder.py

Installation

To get started with recomender, clone this repository and install the required packages using the following commands:

git clone https://github.com/antunsz/recomender.git
cd your-repository-directory
make install

Features

• Distance-Based Reward Building: Utilizes machine learning techniques to calculate rewards based on the distance between entities, improving the efficiency of recommendation systems.
• Modular Design: Easy integration into existing ML workflows, allowing for quick experimentation and deployment.
• Customizable Reward Strategies: Implement your own reward strategies by extending the RewardBuilderAbstract class.

Quick Start

1. Set up your environment

   Ensure you have Python 3.8+ installed. Clone the repository and set up a virtual environment:

   python -m venv venv
   source venv/bin/activate

2. Install dependencies

   Use the Makefile to install the necessary dependencies:

   make install

3. Run the examples

   Check out the examples/ directory to see recomender in action. For a quick start, run the Jupyter notebook:

   jupyter notebook examples/distance_based_reward.ipynb

Development

To contribute to the development of recomender, you can use the following commands:

• Install for development:

  make setup

• Run tests:

  make test

• Linting:

  make lint

For detailed guidelines on contributing, please refer to the CONTRIBUTING.md file.

To Do

• Define the Model Creation Module with support for multiple libraries (Keras, Torch, etc.)
• Implement the RLHF Processing Module
• Create the Reward Generator Module with a Distance-Based Reward Builder
• Implement interaction with Gym Environment for simulation feedback
• Support for External Libraries and Tools in model architecture and training
• Enhance Reward Calculation with additional clustering methods
• Define and implement custom reward strategies

License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Happy recommending! 🎉