TECHTALK
Deep Learning with Bayesian Methods
Tarih:
11/15/2017 11:00 AM
Saat:
YER : Dragos Campus, Library Meeting Room
​​Deep learning is the paradigm that lies at the heart of state-of-the-art machine learning approaches. Despite their groundbreaking success on a wide range of applications, deep neural nets suffer from being severely prone to overfitting and requiring their topology to be heavily handcrafted. Bayesian methods provide principled solutions to both of these problems. Bayesian deep learning converts the loss minimization problem of conventional neural nets into a posterior inference problem by assigning prior distributions on weights. Adaptation of techniques such as variational inference, Markov Chain Monte Carlo sampling, and expectation propagation to deep learning is an attractive research area in the machine learning community. This talk will provide a recap of recent advances in novel Bayesian neural network inference, detail my recent contributions to the solution of this problem, and my ongoing research activities about how Bayesian Neural Nets can be tailored to challenging learning setups, such as active reinforcement learning, few-shot learning, and deep kernel learning.

Speaker:
Assist.Prof. Melih Kandemir
Dr. Kandemir studied computer science in Hacettepe University and Bilkent University between 2001 and 2008. Later on, he pursued his doctoral studies in Aalto University (former Helsinki University of Technology) on the development of machine learning models for mental state inference until 2013. He worked as a postdoctoral researcher in Heidelberg University, Heidelberg Collaboratory for Image Processing (HCI) between 2013 and 2016. He is an assistant professor at Özyeğin University, Computer Science Department since 2017. Throughout his career, he took part in multiple research projects in funded collaboration with multinational corporations including Nokia, Robert Bosch GmbH, and Carl Zeiss AG. Bayesian modeling and inference, weakly supervised learning, active learning, probabilistic deep learning, and application of these approaches to computer vision and medical image analysis problems are among his research interests.