Workshop

The development of artificial intelligence (AI) models is becoming increasingly popular in our industries, having a real impact on our daily lives. The « explainability » and interpretability of AI models are thus questioned today. While in many applications the ramifications of failure are quite low (for example, a movie-recommender fails to provide an accurate prediction), it is imperative that systems where safety is critical, such as automated driving, aerospace, or medical devices, not only get the right answer, but also explain how a final decision was made.
Engineers and scientists must be able to thoroughly understand and investigate a model before putting it into production. In many industries, there are often regulatory requirements to allow a model into production. The higher the ramification of failure, the more need to fully explain the model behavior, and a “black-box” model will not suffice.
Several techniques to improve the explainability of the model will be discussed during this session. This includes simple to more advanced deep learning models with proven visualizations as well as models tested and approved by experts. Emphasis will be placed on the ability to communicate the data properties, decisions and results of the model, detailing not only the inner workings of the model, but how the model was trained, different validation techniques, comparisons with other models, how the data were collected and the importance of training, validation and test sets.
An explainable model is only one part of a complete AI system: a production-ready model must be integrated into a much larger system and may need to be run on specific hardware such as GPUs, FPGAs or the cloud. In this session, using MATLAB, we will lay the foundation for the steps required to implement such a system: from model testing and ease of explanation to system deployment and design.

Speaker: Pierre Harouimi, Application Engineer specialized in data analytics and finance