64th Deep Learning Meetup: DL for Cognitive Strategies, and Image GenAI

Dear Deep Learners,

Our first Deep Learning meetup of the year 2025 will take place on February 20th at Magenta T-Center. We'll have 2 exciting topics:
Deep Learning for Cognitive Strategies, and GenAI for Images in 2025.

***
Agenda:
18:15

• Arrival & registration

18:30

• Introduction by the meetup organizers
• Welcome by the host company

18:45

• Talk 1: The Neural Basis of Cognitive Strategies
  Adam Gosztolai, Medical University of Vienna

19:40

• Announcements
• Networking Break & Discussions

20:15

• Talk 2: GenAI for Images in 2025
  Mouhannad Ali and Julius Duin, Crayon Austria

21:00

• Networking & Discussions

22:00 Wrap up & End

***

Talk Details:
Talk 1: The Neural Basis of Cognitive Strategies
The ability to assess situations, represent tasks flexibly, and choose suitable strategies is central to human intelligence, yet its neural and computational underpinnings remain poorly understood. In the first part of his talk, Adam will describe how automated representation learning techniques can be used to reconstruct the latent dynamical processes generating high-dimensional neural dynamics. He will then highlight a geometric deep learning method, MARBLE, that has specific advantages in discovering latent representations of task-related neural activity and demonstrate its performance on diverse neural decoding tasks. Then, in the second part, the talk will focus on a complex decision-making task combined with large-scale electrophysiology, designed to probe the rich space of cognitive solutions in macaques. Extensive neural decoding showed that macaques inductively generalise a small set of cognitive strategies to a vast space of tasks. Finally, the speaker will demonstrate how MARBLE representations revealed the dynamical configurations that allow monkeys to plan and implement distinct cognitive heuristics. These neurally inferred strategies provide an understanding of the neural basis of adaptive problem-solving.

About the speaker:
Adam Gosztolai is an Assistant Professor for AI and Neuroscience and a research group leader of the “Dynamics of Neural Systems Laboratory” at Medical University of Vienna and a research affiliate at the Department of Cognitive Sciences at MIT. He read engineering and mathematics at University College London and the University of Cambridge and obtained his PhD in mathematics from Imperial College London. Following his PhD, Adam has done postdoctoral research at the École Polytechnique Férédale de Lausanne, in the fields of computational neuroscience and machine learning. For his research he was awarded a Human Frontiers Science Program Fellowship and ERC Starting grant and a WWTF Vienna Research Group grant. In his research, Adam studies the dynamical processes encoded in the activity of a large number of neurons in the brain to distil fundamental principles of how these collective dynamics are linked to neural processes such as cognition and motor control. He aims at transferring these principles to design artificial neural systems that mimic how the brain works.

Talk 2: GenAI for Images in 2025
Remember when AI-generated images were just weird, distorted curiosities? Fast forward to today, and we're witnessing nothing short of a creative revolution. In this session, we'll jump into the world of text-to-image generation that has exploded in 2024, turning imagination into pixels at unprecendented speed.
Get ready for an exploration of:

• Breakthroughs that transformed platforms like Midjourney, DALL-E, and Flux from "interesting tools" into "creative superpowers"
• Behind-the-scenes looks at the magic sauce: the architectural innovations that made all this possible
• A peek into the crystal ball: what's cooking for 2025?

Note: This session assumes familiarity with deep learning fundamentals and transformer architectures.
About the speakers:
Mouhannad Ali, PhD, Senior ML Engineer at Crayon Austria, is an AI and machine learning expert with over 15 years of experience in academia and industry. His work spans transportation, healthcare, and financial services, focusing on AI/ML applications and product development.
With a strong track record of innovation, he has led impactful projects in both research and industry, solving complex real-world challenges with advanced data science techniques.
Julius Duin, ML Engineer at Crayon Austria, started their journey in psychology and computer science before diving into neuroscience, bringing a unique perspective to technology and human behavior.
After co-founding a fitness tech startup focused on motion tracking in gyms, he discovered another passion at the intersection of computer vision and machine learning.
Today, he’s an ML engineer who's particularly fascinated by computer vision and robotics, drawing interesting parallels between artificial systems and their neuroscience background.

We are really looking forward to enlightening talks and discussions, and would like to thank Magenta for providing drinks & snacks at this meetup.

Your VDLM organizers
Tom, René, Pavol, Jan and Alex