63rd Deep Learning Meetup: LLMs & Mental Health / Robust Multimodal Learning

Dear Deep Learners,

We kindly invite you to our next Deep Learning meetup on November 7th at FH Technikum Wien. We'll have 2 exciting topics:
LLMs & Mental Health and Robust Multimodal Learning.

***

Agenda

18:30

• Introduction by the meetup organizers
• Welcome by FH Technikum: Rafael Rasinger and Isabel Dregely

18:45

• Case Study: When LLMs Meet the Clinical Trial
  Adam Kolář

19:40

• Announcements
• Networking Break & Discussions

20:15

• Robust Multimodal Learning
  Shah Nawaz, JKU Linz
• Networking & Discussions

21:30 Wrap up & End

***

Talk Details:

Talk 1: Case Study: When LLMs Meet the Clinical Trial
As the hype around large language models (LLMs) has grown, so has the open-source toolbox, offering surprising capabilities even with limited budgets and computational resources. In this case study, we’ll present practical and easily transferable methods for applying small (8b-class) LLMs to a unique dataset from clinical trials for anxiety treatments. Our focus is on minimizing human-factor variability in measuring anxiety using the Hamilton Anxiety Scale (HAM-A). We’ll demonstrate how language models can automate these assessments from HAM-A interview audio recordings, identify inconsistencies, and provide feedback to clinicians.

About the speaker: For over a decade, Adam Kolář has contributed to a wide array of AI projects across business and academia. His journey includes developing image search and enhancing full-text search at Seznam.cz, researching general artificial intelligence at GoodAI, achieving precise measurements in healthcare with HealthMode, and founding startups—some successful, some not—for VC investors. Beyond his professional roles, Adam is an educator, offering courses and mentoring companies. He is also a founding member of the Brno Machine Learning Meetups.

Talk 2: Robust Multimodal Learning
Multimodal networks have demonstrated remarkable performance improvements over their unimodal counterparts. Existing multimodal networks are designed in a multi-branch fashion that, due to the reliance on fusion strategies, exhibit deteriorated performance if one or more modalities are missing.
In this work, we will discuss multimodal learning methods inherently robust to missing modalities.

Hot Topics & Latest News:
Do you have some interesting breaking news about Deep Learning? Did you read an interesting paper that you want to share? Did you create an exciting application or achieve some break-through? It would be great to share this in our meetup's Hot Topics section! Please get in touch through contact@vdlm.at

We are looking forward to seeing you at our next meetup!
FH Technikum will kindly provide drinks & snacks.