Norwegian version of this page

dScience Lunch Seminar: Balancing Explainability-Accuracy of Complex Models

This week’s dScience lunch seminar delves into one of AI’s pressing challenges: how to balance explainability and accuracy in complex machine learning models. Poushali Sengupta, Ph.D. Research Employee at the Department of Informatics, will introduce ExCIR (Explainability through Correlation Impact Ratio), a novel method that addresses this trade-off.

Image may contain: Font, Electric blue, Magenta, Parallel, Logo.

Presentation

Explainability in AI is crucial across various applications, yet existing methods often overlook the explainability-accuracy trade-off in complex models. We introduce Explainability through Correlation Impact Ratio (ExCIR), a novel approach that balances explainability and accuracy in complex machine learning models, especially with independent features. ExCIR creates a streamlined data space from the original dataset, requiring fewer input samples while maintaining model accuracy. The Correlation Impact Ratio (CIR) quantifies each feature’s contribution to the model’s output, accounting for feature uncertainty regarding its presence or absence. Using the CAUEEG dataset related to dementia, we demonstrate that ExCIR maintains accuracy while enhancing explainability through feature impact scores. Our method reliably ranks feature importance in original and simplified models, validating its consistency.

Speaker

Poushali Sengupta joined the Department of Informatics as a Ph.D. Research Employee in November 2021. Her research area is related to explainable artificial intelligence, distributed machine learning, data privacy, and security. She is currently working on Explainable AI to find a balance in an accuracy-explainability tradeoff for complex ML models.

Program

11:30 – Doors open and lunch is served

12:00 – "Balancing Explainability-Accuracy of Complex Models" by Poushali Sengupta (PhD Candidate, Department of Informatics)

This event is open for all students, PhD candidates, postdocs, and everyone else who is interested in the topic. No registration needed.

About the seminar series

Once a month, dScience will invite you to join us for lunch and professional talks at the Science Library. In addition to these, we will serve lunch in our lounge in Kristine Bonnevies house every Thursday. Due to limited space (40 people), this will be first come, first served. See how to find us here.

Our lounge can also be booked by PhDs and Postdocs on a regular basis, whether it is for a meeting or just to hang out – we have fresh coffee all day long!

Lounge Calendar

Tags: dscience, postdoc, phd, lunch seminar
Published Aug. 9, 2024 11:47 AM - Last modified Nov. 11, 2024 1:35 PM