This repository provides implementations of machine learning and AI algorithms for medicine, developed by the van der Schaar Lab for papers published before 2023. It serves researchers and practitioners in healthcare AI, offering a collection of advanced methods for tasks like treatment effect estimation, survival analysis, and data synthesis.

How It Works

The repository is structured into directories for algorithms (alg/), applications (app/), configuration (cfg/), documentation (doc/), and utilities (util/). It houses implementations of diverse ML techniques, including Gaussian Processes, Generative Adversarial Networks (GANs), deep learning models, and reinforcement learning, often tailored for complex medical data and causal inference problems. The approach emphasizes novel architectures and theoretical underpinnings for robust medical decision-making.

Quick Start & Requirements

Installation and usage instructions are specific to each algorithm and application directory. Common installation guidance is available in doc/install.md. Data files and models can be found via DATA-*.md files in respective directories.

Highlighted Details

• Extensive collection of algorithms for causal inference, survival analysis, time-series modeling, and privacy-preserving data synthesis.
• Includes implementations for notable applications like Adjutorium COVID-19 and Clairvoyance for medical time series.
• Detailed mapping of publications to code locations, facilitating research reproducibility.
• Covers a wide range of conferences and journals including NeurIPS, ICML, ICLR, and AAAI.

Maintenance & Community

This repository contains legacy code from before 2023; newer work is available at https://github.com/vanderschaarlab. Specific community or maintenance details are not provided in the README.

Licensing & Compatibility

The software is generally released under the 3-Clause BSD license, with exceptions noted for individual algorithms and applications. This license is permissive and generally compatible with commercial use.

Limitations & Caveats

This repository contains code for papers published before 2023, indicating it may not reflect the lab's most current research or methodologies. Installation and dependency management can vary significantly between individual algorithm/app directories.