IDL RESEARCH SUMMER SCHOOL 2024

The course is available for PhD fellows with scholarship and industry professionals.

*Application deadline: April15th, 2024

The course is limited to 25 places.

INF-8605-1 24H IDL (2024) Course Lectures

Interpretability in
Deep Learning

Welcome to the summer school “Interpretability in Deep Learning 2024”, organized by the DLN center at UiT The Arctic University of Norway!

Oncampus Summer School: 7th - 9th August, 2024

Venue : TEKNOBYGGET 1,022AUD, Tromsø, Norway

Application Link: https://uit.no/utdanning/emner/emne/842308/inf-8605

Dedicated Course Page: https://www.indeeplearning.org/course

Artificial intelligence (AI) and machine learning approaches are often considered as black boxes, i.e. as a type of algorithms that accomplish learning tasks but cannot explain their knowledge. However, as AI is increasingly adopted for accomplishing human cognitive tasks, it is becoming important that the AI models are understandable. In critical tasks such as deriving, from given data, a correct medical diagnosis and prognosis, collaboration between artificial and human intelligence in imperative so that the suggestions or decision from AI are both more accurate and more trustworthy. This intensive course will consider different topics of importance regarding explainable AI, equipping the students with knowledge of approaches that can be used to interpret AI, and AI approaches that are more explainable than others. In addition, the students will receive practical skills of applying selected approaches, which will equip the students with practical skills of adapting to the rapid pace of technology development in the field of explainable AI.

Introductory Concept

Unveiling the Black-box Problem
Tracing the Evolution of Interpretability and Explainability
Knowledge versus Performance, Need of Explainability
Overview of Course & Learning Resources

3 hours

Explainable Approaches

Understanding Knowledge Encoding, Feature Importance and Feature Interactions
Neural Network Perturbation: Role of Neurons, LIME, Occlusion
Theory and Intuition behind CAM
Neural Network Gradient: CAM, Grad-CAM, Saliency, LRP.

6 hours

Model-agnostic Techniques

Abstract Encoding Concept
Case Study: Visual Propagation-based Explanations
Sanity Check and Evaluation Criteria – Randomization Test
Introduction to Adversarial Attacks and Defenses -
Counterfactual Reasoning and Causal Inference Models
Bias Estimation and Mitigation

5 hours

Neural Network and Explainability

Segregating methods into the new class of representation
Harnessing the Power of Self-interpretable Models: Towards Robust Interpretability
Interpretability Direction Towards Sustainable AI
Understanding the nature of bio-medical microscopy application

5 hours

Ethics & Emerging Trend in AI

Ethics in Artificial Intelligence
Diverse scenarios and impact on industrial standards
Fairness, accountability, and transparency in deep learning
Review of GDPR regulation
Emerging trends in explainable AI

1 hours

NOTE: First introductory lecture was held on 28th May 2024 (online). [YouTube]

Part of the lectures will be uploaded as video lectures, which is a prerequisite before the onsite (at UiT) program begins in August. The details have been shared during the first digital introductory.