Human-Centric Artificial Intelligence projects

This repository contains Django implementations of Human-Centric Artificial Intelligence projects:

---

Project Structure

The project is organized into 5 main parts:

1. Part 1: Automated Machine Learning - Supervised Learning Interface

   • Purpose: Provide a web-based interface for supervised learning tasks, enabling users to uplaod datasets, visualize them, and train machine learning models interactively.
   • Key features:
     • 📂 CSV dataset upload
     • 📊 Visualization of features and labels
     • ⚙️ Model training with train/test split, hyperparameter tuning, and evaluation
     • 🔄 User choice between different algorithms and settings
     • 🖥 Integration with Django interface

2. Part 2: Active Learning for Texr Classification

   • Purpose: Develop an interactive text classification system for seniment analysis on the IMDB 50k dataset, with a focus on active learning. Instead of relying on a large labeled dataset, the system incrementally queries the user (or a simulated annotator) to label the most informative samples, reducing annotation effort while maintaing high performance.
   • Key features:
     • 📝 Train a supervised text classifier (baseline accuracy).
     • 🎯 Implement pool-based active learning strategies.
     • 👩‍💻 Support simulated or real users for labeling.
     • 📈 Track model performance as labeling progresses.
     • 🔄 Optional extensions: batch or stream-based active learning.

3. Part 3: Explainability

   • Purpose: Provide an interactive tool for exploring model interpretability and counterfactual explanations using the Palmer Penduins dataset. The app highlights the trade-off between accuracy and complexity while helping users understand and question model predictions.
   • Key features:
     • 🌳 Train interpretable models (decision trees, logistic regression) with adjustable complexity.
     • 🎚 Interactive slider to control sparsity vs. accuracy.
     • 📊 Display accuracy, model size (leaves/features), and structure.
     • 🔄 Generate counterfactual explanations for selected samples.
     • 👩‍💻 User interface for exploring predictions and alternatives.

4. Part 4: Influence of future predictions over active learning of users' tastes for recommender systems

   • Purpose: Design an interactive recoomender system that tackles the cold start problem by learning user preferences through guided input and showing how their ratings affect recommendations.
   • Key features:
     • 🎬 Use the MovieLens dataset for movie recommendations.
     • 📝 Guided active learning: users rate movies while seeing how their input impacts the system.
     • 👩‍🔬 Design and document a user study to test hypotheses about the method.
     • 🖥 Build an interactive interface for the study with a landing page, PDF documentation, and study launch.
     • ⚙️ Implement recommendations using matrix factorization with user/item embeddings.

5. Part 5: Reinforcement Learning with Human Feedback

   • Purpose: Develop a reinforcement learning system where a mouse learns to navigate a grid-world and collect cheese, refined by human feedback to prefer desirable outcomes and avoid traps.
   • Key features:
     • 🐭 Grid-world game environment with cheese, traps, and walls.
     • ⚡ Learn policies using the REINFORCE algorithm and neural networks.
     • 👩‍💻 Incorporate human feedback (or simulated feedback) to refine the learned policy.
     • 📊 Estimate rewards from feedback using the Bradley-Terry model.
     • 🔄 Retrain the policy network with the refined reward while maintaining baseline behavior.

---

⚙️ Installation

Requirements

• Install Docker and Docker Compose on your system
  • Docker Desktop for Mac/Windows: https://www.docker.com/products/docker-desktop/
  • Linux: sudo apt install docker.io docker-compose

Steps

1. Clone the repository: git clone https://github.com/sevenxx07/Human-CentricAI_Project.git

2. Build the Docker image: docker-compose build

   • This uses the Dockerfile in docker/Dockerfile to create an environment with all dependencies

3. Run the project: docker-compose up

   • The command automatically runs:
     • python manage.py makemigrations
     • python manage.py migrate
     • python manage.py runserver 0.0.0.0:8000

• Your Django app will be accessible at: http://127.0.0.1:8000

4. Stop the project: docker-compose down

Windows specific settings:

If you are running the docker on Windows follow also these additional steps:

1. docker-compose.yml - remove line 6 network: host

2. make sure your windows firewall is not blocking proper docker composing