| --- |
| license: apache-2.0 |
| base_model: |
| - ncbi/MedCPT-Query-Encoder |
| tags: |
| - federated-learning |
| - blockchain |
| - healthcare |
| - clinical-nlp |
| - semantic-search |
| - privacy-preserving |
| pipeline_tag: feature-extraction |
| --- |
| |
| # Global Model: Decentralized Federated Learning for Distributed Hospitals |
|
|
| This repository hosts the global model weights aggregated during the decentralized federated learning process across distributed clinical institutions. The aggregation coordination, model integrity verification, and round participation are orchestrated via a blockchain network to ensure auditability, zero data leakage, and Byzantine fault tolerance. |
|
|
| ## Model Details |
|
|
| ### Model Description |
|
|
| - **Model type:** Biomedical Query Encoder / Text Feature Extraction |
| - **Language(s) (NLP):** English |
| - **License:** Apache-2.0 |
| - **Finetuned from model:** `ncbi/MedCPT-Query-Encoder` |
| - **Federated Scheme:** Decentralized On-Chain Aggregation (e.g., FedAvg / Smart Contract Consensus) |
|
|
| ### Model Sources |
|
|
| - **Repository:** [GitHub: Decentralized-Systems-and-Applications/Federated_Learning_with_Blockchain_for_Distributed_Hospitals](https://github.com/Decentralized-Systems-and-Applications/Federated_Learning_with_Blockchain_for_Distributed_Hospitals) |
| - **Demo :** [ Demo Video (https://drive.google.com/file/d/11C0rO3Zjp7459Yn7012KK0VBu-refau_/view)] |
| |
| --- |
| |
| ## Uses |
| |
| ### Direct Use |
| Generating dense clinical embeddings for biomedical text retrieval, semantic document matching, and hospital query classification without exposing institutional private training data. |
| |
| ### Downstream Use |
| Integrating into distributed clinical decision support systems (CDSS), electronic health record (EHR) cross-search, or zero-trust medical intelligence pipelines. |
| |
| ### Out-of-Scope Use |
| - Automated diagnosis or medication prescribing without human clinician oversight. |
| - Handling tasks completely outside the clinical and biomedical domain without domain-specific re-evaluation. |
| |
| --- |
