How lavanya212005/adaptive-federated-learning-for-non-iid-healthcare-data works
This repository combines two breast-ultrasound workflows: a Streamlit prediction/history/reporting app and a federated-learning research pipeline. The app’s sampled prediction helper generates filename-biased random probabilities rather than invoking the neural models, so the two workflows are not presented as connected. Federated learning prepares client data, trains models with several strategies, and records metrics; sampled excerpts do not establish the full orchestration links. The graph preserves the documented app and analysis capabilities while omitting uncertain connections.
Main entry points: predict.py Image prediction · model.py Training model and strategy · app.py Streamlit interface · data_utils.py Client data splits
Components 14
Prediction App
- Streamlit interface
app.py - Image prediction
predict.py - Prediction helpers
utils.py - Prediction history
utils.py - PDF reports
utils.py - Heterogeneity UI
newapp.py
Federated Learning
- Client data splits
data_utils.py - Training configuration
config.py - CNN model
models.py - Training model and strategy
model.py - Federated utilities
fl_utils.py - Training metrics
fl_utils.py
Dataset Analysis
- Dataset analysis
analyse.py
Other components
- Patient or researcher
Connections 13
- Patient or researcher → Streamlit interface: uses
app.py - Patient or researcher → Image prediction: uploads images
predict.py - Streamlit interface → Image prediction: renders tab
- Image prediction → Prediction helpers: generates predictions
predict.py - Image prediction → Prediction history: saves entries
predict.py - Image prediction → PDF reports: builds report
predict.py - Patient or researcher → Heterogeneity UI: uses
newapp.py - Training configuration → Client data splits: configures splits
data_utils.py - Client data splits → Training model and strategy: returns client data
- Training model and strategy → CNN model: trains model
- Training model and strategy → Federated utilities: uses strategy helpers
- Federated utilities → Training metrics: writes metrics
fl_utils.py - Patient or researcher → Dataset analysis: runs analysis