Tech talk

Technical Details of Diffusion-Based Generative Models

Technical Details of Diffusion-Based Generative Models

 

Tech Talk with Ahmet Gündüz

Bio

Ahmet Gündüz received his B.Sc. degree in Electrical and Electronics Engineering from Boğaziçi University, Istanbul, one of the top universities in Turkey. He holds two M.Sc. degrees, both from top German Universities: the first in Telecommunication Engineering from Technical University Munich and the second in Data Science from Maximilian University of Munich Ludwig, Germany. He worked on Gesture Recognition Technologies in his master’s thesis and got the best student paper award at IEEE FG 2019 conference with his work. He is actively continuing his research in various fields, such as computer vision, speech processing and NLP, for aiXplain Inc. as AI/ML Researcher.

Abstract

Ahmet Gündüz covers the role of AI / ML Researcher at aiXplain.
In this talk, Ahmet will briefly introduce some generative models like Variational Autoencoders (VAEs), Generative Adversarial Networks (GANs), and Flow-Based Generative Models.
He will also give technical details about Diffusion Models starting from Markov Chain to Inverse Diffusion Process.
The talk will get more technical from this point, discussing the advantages of Diffusion Models over other generative models. In the end, Ahmet will demonstrate image generation through a Colab notebook.

Registration is closed.