Federated Learning for Chemical Safety Assessment

This repository contains three distinct federated learning implementations for hazard and exposure chemical safety assessment.

🧪 Projects

Federated Analytics

• Model: Histogram aggregation
• Task: Dermal permeability (logKp) analysis
• Data: HuskinDB + SkinPiX + NCSU_USEPA datasets
• Approach: Privacy-preserving analytics with differential privacy

Horizontal Federated Learning

• Model: XGBoost
• Task: Skin sensitisation prediction (LLNA)
• Data: AI4Cosmetics + SkinDoctorCP datasets
• Approach: Organisations with different datasets collaborate

Vertical Federated Learning

• Model: SMILES CNN
• Task: Mutagenicity prediction
• Data: Non-IID molecular data splits
• Approach: Clients with different data distributions collaborate

🚀 Quick Start

Each project is self-contained with its own setup and usage instructions:

# For federated analytics (Histograms)
cd federated_analytics/
# Follow federated_analytics/README.md

# For horizontal FL (XGBoost)
cd horizontal_fl/
# Follow horizontal_fl/README.md

# For vertical FL (SMILES CNN)  
cd vertical_fl/
# Follow vertical_fl/README.md

📁 Repository Structure

fl-chemsafe/
├── federated_analytics/                     # Privacy-preserving histogram aggregation
│   ├── README.md                            # Setup and usage instructions
│   ├── federated_analytics/                 # Federated learning implementation
│   │   ├── client_app.py                    # Flower client implementation
│   │   ├── server_app.py                    # Flower server implementation  
│   │   └── task.py                          # Data loading and model logic
│   ├── pyproject.toml                       # Python dependencies
│   ├── data/                                # Datasets
│   └── plots/                               # Generated visualisations
├── horizontal_fl/                           # XGBoost federated learning
│   ├── README.md                            # Setup and usage instructions
│   ├── horizontal_fl/                       # Federated learning implementation
│   │   ├── client_app.py                    # Flower client implementation
│   │   ├── server_app.py                    # Flower server implementation
│   │   └── task.py                          # XGBoost model logic
│   ├── data_preparation.py                  # Dataset preprocessing
│   ├── baseline_models_evaluation.py        # Performance comparison
│   ├── create_performance_plot.py           # Performance visualisation
│   ├── create_umap_plot.py                  # UMAP embedding plots
│   ├── pyproject.toml                       # Python dependencies
│   ├── data/                                # Datasets
│   ├── models/                              # Saved model files
│   ├── results/                             # Evaluation results
│   └── plots/                               # Generated visualisations
├── vertical_fl/                             # SMILES CNN federated learning
│   ├── README.md                            # Setup and usage instructions  
│   ├── vertical_fl/                         # Federated learning implementation
│   │   ├── client_app.py                    # Flower client implementation
│   │   ├── server_app.py                    # Flower server implementation
│   │   └── task.py                          # CNN model and data logic
│   ├── data_preparation.py                  # Non-IID data splits
│   ├── evaluate_local_models.py             # Local model evaluation
│   ├── create_performance_visualization.py   # Performance plots
│   ├── pyproject.toml                       # Python dependencies
│   ├── data/                                # Datasets
│   ├── results/                             # Evaluation results
│   └── plots/                               # Generated visualisations
└── README.md                                # This file