Mentorship Talk for WiMLDS and Research Talk on Deep Learning System Evaluation

Join us on July 23rd for an evening of insightful talks and networking opportunities.

This event is perfect for the curious! Are you thinking about your career in the era of Large Language Models? Or are you eager to learn more about deep learning systems, their evaluation, and applications? You are all most welcome to join us this evening.

The event will begin with an interactive mentorship talk by Alessandra Zarcone, a Professor of Language Technology and Cognitive Assistance at the Faculty of Computer Science of the Technische Hochschule Augsburg, followed by a technical presentation by Sheethal Bhat, a PhD candidate at Siemens Healthineers in Erlangen. We encourage the audience to actively engage in the discussions in both sessions.

Following the event, we plan to open a casual networking opportunity. More updates will be provided soon by Email.

All genders are cordially invited to attend this event.

Please RSVP on Meetup if you are planning to join and adjust your Meetup settings to receive Emails about the event update.

Note: This is a hybrid event. Those participants who RSVP on the Meetup will receive a Zoom link a day before the event.

We are looking forward to seeing you soon!

Date & Time:
Tuesday, 23.07.2024, 17:30 - 19:30

Location:
Medical Valley Center, Henkestr. 91, 91052 Erlangen,
Room K2, 2nd floor

Agenda:
17:20 Arrival
17:30 Introduction to WiMLDS activities
17:40 Interactive mentorship talk: The Importance of Being Hybrid: Research and Career in the Time of Large Language Models by Alessandra Zarcone
18:50 Break
19:00 Presentation: AUCReshaping: improved sensitivity at high-specificity by Sheethal Bhat
19:30 Closing, mingle, and network

About presentations:

Mentorship Talk:
The recent developments in LLMs and Generative AI have brought visible changes in the fields of DS, ML, AI and NLP. Business models have changed, new skills are now require, new professional figures have emerged, and the potential for discrimination has increased. In this mentorship talk, I will focus in particular on what this means for women in the fields of DS, ML, AI and NLP, from the point of view of the technological advancements as well as from the point of view of cultural shifts and changes in the scientific community and in society.

Technical Presentation:
The evaluation of deep-learning (DL) systems typically relies on the Area under the Receiver-Operating-Curve (AU-ROC) as a performance metric. However, AU-ROC, in its holistic form, does not sufficiently consider performance within specific ranges of sensitivity and specificity, which are critical for the intended operational context of the system. Consequently, two systems with identical AU-ROC values can exhibit significantly divergent real-world performance. This issue is particularly pronounced in the context of anomaly detection tasks, a commonly employed application of DL systems across various research domains, including medical imaging, among others. The challenge arises from the heavy class imbalance in training datasets, with the abnormality class often incurring a considerably higher misclassification cost compared to the normal class. Traditional DL systems address this by adjusting the weighting of the cost function or optimizing for specific points along the ROC curve. While these approaches yield reasonable results in many cases, they do not actively seek to maximize performance for the desired operating point. In this talk, we introduce AUCReshaping, designed to reshape the ROC curve exclusively within the specified sensitivity and specificity range, by optimizing sensitivity at a predetermined specificity level. This reshaping is achieved through an adaptive and iterative boosting mechanism that allows the network to focus on pertinent samples during the learning process. We primarily investigated the impact of AUCReshaping in the context of abnormality detection tasks, specifically in Chest X-Ray (CXR) analysis, followed by breast mammogram and credit card fraud detection tasks. The results reveal a substantial improvement, ranging from 2 to 40%, in sensitivity at high-specificity levels for binary classification tasks.

Code of Conduct:
Our WiMLDS meetups aim to inspire, educate, and support women and gender minorities. We are inclusive to anyone who supports our cause regardless of gender identity or technical background. All genders are cordially invited to attend this event.

WiMLDS is dedicated to providing a harassment-free experience for everyone. We do not tolerate harassment of participants in any form. All communications should be appropriate for a professional audience including people of many different backgrounds. Sexual language and imagery are not appropriate.
All attendees should read the full Code of Conduct before participating: https://github.com/WiMLDS/starter-kit/wiki/Code-of-conduct