Responsible deployment of deep learning

Program Overview

This 3-day program provides an overview of the interpretability, deployment, and responsible use of deep learning models. Participants will learn about the concepts of interpretability, deployment, and the social and ethical implications of deep learning. The program includes hands-on exercises using Python libraries such as SHAP and LIME, and frameworks such as Strealit and FastAPI.

Program Outline

Outline:

Day 1: Interpretability of artificial neural networks

• How to understand and explain what happens in an artificial neural network?
• Modeling in the form of a “black box”.
• The main concepts covered in this module are: interpretability, action plan and policy of a model, and the Python SHAP and LIME libraries.

Day 2: Deployment of a deep learning model

• Concrete example of the deployment of a deep learning and focus model on a few specialized frameworks.
• The main concepts covered in this module are: deployment, data science workload and Strealit and FastAPI frameworks.

Day 3: Responsible deployment of a deep learning model

Teaching:

TRAINERS / TEACHERS

• Emmanuel R. Goffi
• Virginie Mathivet