Study Heart Disease using PySyft

What you will learn 🎓

1. work remotely with non-public medical data.
2. use PySyft to run Machine learning experiments on non-public and distributed medical datasets.
3. take advantage of getting access to multiple medical datasets for better Machine learning modelling.

Materials 🧑‍💻

The tutorial is organised into multiple Jupyter notebooks that will guide you to the different steps of our Machine learning experiment, using PySyft.

• 🧭 (Intro) Setup Datasites

  This introductory notebook will help you to get your bearings with the data, and the PySyft Datasites.

• 📊 1. Compare Demographics

  In this notebook we will start our data science journey, analysing the distribution of demographics in the data. Why this is interesting ? We will learn how to get our first insights about the data, without ever seeing the data!

• 🤖 2. ML Model Training Experiment:

  Let's use PySyft to train a Machine learning classifier, using data across the four distributed datasites, while also maintain the non-public data private! (This is going to be our 🌟 research experiment with PySyft!)

• 📝 3. ML Model Evaluation Experiment:

  It's now time to assess the performance of the trained classifiers on each remote datasite. We will gather evaluation metrics for the trained models, and compare the results! We will learn how to create specialised Syft function that guarantees control over the input/output policies of our remote code execution. (This will be our 🌟🌟 research experiment with PySyft!)

• 🗳️ 4. Ensemble Learning Experiment :

  In this last step, we will use an Ensemble as a strategy to combine the multiple ML models trained on the four medical datasets. In this way, we will be able to use a ML model that has seen 4x more data to generate our heart disease prediction! (This will be our 🌟🌟🌟 and last research experiment with PySyft!)