Explainable AI

Define XAI Taxonomy
Distinguish between Antehoc and Posthoc explainability models
Application of XAI methods in Deep Learning Models
Estimate the performance of different XAI methods
Requirements
Prior knowledge on Deep Learning and Machine Learning Models
Description
This course is designed to provide a sound understanding on the application of various explainable AI methods for DL models and the intrinsically interpretable aspects of ML models.
Intelligence integrated Data Analytics utilized for Real time systems under different problem settings, helps to take immediate decisions with highest level of accuracy. In order to ensure this, it is necessary to prove the accuracy of analytics algorithms based on the data captured and Machine Learning algorithms applied on these applications. Calculating accuracy, precision, recall and F1 score may help in selecting the appropriate Machine Learning or Deep Learning model for prediction. However, as AI based analytics systems equipped with ML / DL algorithms are generally black boxes, where users can see only the inputs and outputs but not the inner workings. There are no proper evidences that the prediction given by these systems is correct. This lack of transparency makes it inherently challenging to understand how and why they reach their decisions. For human operators or end users to trust AI systems, these explanations must align with human perception and expectations. Yet current machine learning or deep learning models tend to be untrustworthy due to their inherent complexity and lack of transparency.
Who this course is for
UG and PG students, Researchers, Industriies
Homepage