Qualcomm-UvA Deep Vision Seminars

New challenges in Reinforcement Learning

Oriol Vinyals
Research Scientist, Google

Date:
September 15, 14:00-15:00

Location:
University of Amsterdam
Room C0.110
Science Park 904
1098 XH Amsterdam

Abstract:
In this talk, Oriol will describe some of the recent work that he and his collaborators have done at DeepMind in model based RL, and their recent work on StarCraft II, a strategy game which poses a new challenge for deep RL.

Bio:
Oriol Vinyals is a Staff Research Scientist at Google DeepMind, working in Deep Learning. Prior to joining DeepMind, Oriol was part of the Google Brain team. He holds a Ph.D. in EECS from University of California, Berkeley and is a recipient of the 2016 MIT TR35 innovator award. At DeepMind he continues working on his areas of interest, which include artificial intelligence, with particular emphasis on sequences, deep learning and reinforcement learning.

Website: https://deepmind.com/blog/deepmind-and-blizzard-release-starcraft-ii-ai-research-environment/

Modeling Human Perception with Deep Neural Networks

Prof. M. Bethge
University of Tübingen

Date:
March 10, 2017, 15:00-16:00

Location:
University of Amsterdam
Room C0.110
Science Park 904
1098 XH Amsterdam

Abstract:
Last year, we introduced Qualcomm-UvA Deep Vision Seminars, hosted by Prof. I. Kokkinos from INRIA Saclay, Dr. Max Jaderberg from Google Deepmind, and Dr. J. Yosinksi from Geometric Intelligence. The goal of the seminars was to invite seminal guest speakers to provide talks on the latest advances in the areas of deep learning, computer vision, and machine learning. In our next seminar, we will have a talk by Prof. M. Bethge, whose work on Neural Artistic Style Transfer you may have seen quite extensively in the media for making your photograph look like a "van Gogh".

Bio:
Prof. M. Bethge did his undergraduate studies in physics and started working in computational neuroscience when he joined the MPI. Since then, his research continues to aim at understanding perceptual inference and self-organized collective information processing in distributed systems---two puzzling phenomena that contribute much to our fascination about living systems. He believes that general principles are important, but at the same time, these principles need to be grounded in reality. Therefore, a large part of his research focuses on the mammalian visual system working closely together with experimentalists (e.g. Andreas Tolias, Thomas Euler, and Felix Wichmann). Additionally, he works on neural coding in other sensory systems, collaborating with Cornelius Schwarz. 

Website: http://bethgelab.org/people/matthias/

A Deeper Understanding of Large Neural Nets

Dr. Jason Yosinksi
Geometric Intelligence

Date: December 2, 2016, 14:00-15:00

Location:
University of Amsterdam
Room C1.110
Science Park 904
1098 XH Amsterdam

Abstract:
Deep neural networks have recently been making a bit of a splash, enabling machines to learn to solve problems that had previously been easy for humans but hard for machines, like playing Atari games or identifying lions or jaguars in photos. But how do these neural nets actually work? What do they learn? This turns out to be a surprisingly tricky question to answer - surprising because we built the networks, but tricky because they are so large and have many millions of connections that effect complex and hard to interpret computation. Trickiness notwithstanding, in this talk we’ll see what we can learn about neural nets by looking at a few examples of networks in action and experiments designed to elucidate network behavior. The combined experiments yield a better understanding of network behavior and capabilities and promise to bolster our ability to apply neural nets as components in real world computer vision systems.

Bio:
Jason Yosinski is a researcher at Geometric Intelligence, where he uses neural networks and machine learning to build better AI. He was previously a PhD student and NASA Space Technology Research Fellow working at the Cornell Creative Machines Lab, the University of Montreal, the Caltech Jet Propulsion Laboratory, and Google DeepMind. His work on AI has been featured on NPR, Fast Company, Wired, the Economist, TEDx, and BBC.

Website:  yosinski.com

Related papers: yosinski.com/papers

Temporal Credit Assignment for Training Recurrent Neural Networks

Date:
October 28, 2016, 11:30-13:00

Speaker:
Max Jaderberg
Google DeepMind

Location:
University of Amsterdam, Science Park
Building: Science Park 904
Room: C0.110

Abstract:
The problem of temporal credit assignment is at the heart of training temporal models -- how the processing or actions performed in the past affects the future, and how we can train this processing to optimise future performance. This talk will focus on two distinct scenarios. First the reinforcement learning scenario, where we consider an agent which is a recurrent neural network which takes actions in its environment. I will show our state-of-the-art approach to deep reinforcement learning, and some of the latest methods which deal with enhancing temporal credit assignment, presenting results on new 3D environments. I will then look at how temporal credit assignment is performed more generically during the training of recurrent neural networks, and how this can be improved by the introduction of Synthetic Gradients -- predicted gradients from future processing by local models learnt online.

Bio:
Max Jaderberg is a Senior Research Scientist at Google DeepMind, working on machine learning, reinforcement learning, and computer vision. He previously co-founded Vision Factory which was acquired in 2014 by Google, and completed his PhD at the University of Oxford under the supervision of Andrew Zisserman and Andrea Vedaldi in the Visual Geometry Group, where his main focus of research was text recognition and detection using deep neural networks.

Deeplab to UberNet: from task-specific to task-agnostic deep learning in computer vision

Date:
September 26, 2016, 13:30-15:00

Speaker:
Iasonas Kokkinos
CentraleSupelec/INRIA

Location: 
University of Amsterdam
Room C3.163
Science Park 904
1098 XH Amsterdam

Abstract:
Over the last few years Convolutional Neural Networks (CNNs) have been shown to deliver excellent results in a broad range of low- and high-level vision tasks, spanning effectively the whole spectrum of computer vision problems.
In this talk we will present recent research progress along two complementary directions.

In the first part we will present research efforts on integrating established computer vision ideas with CNNs, thereby allowing us to incorporate  task-specific domain knowledge in CNNs. We will present CNN-based adaptations of structured prediction techniques that use discrete (DenseCRF - Deeplab) and continuous energy-based formulations (Deep Gaussian CRF), and will also present methods to incorporate ideas from multi-scale processing, Multiple-Instance Learning and Spectral Clustering into CNNs.

In the second part of the talk we will turn to designing a generic architecture that can tackle a multitude of tasks jointly, aiming at designing a `swiss knife’ for vision. We call this network an ‘UberNet’ to underline its overarching nature. We will introduce  techniques that allow us to train an UberNet while using datasets with diverse annotations, while also handling the memory limitations of current hardware. The proposed architecture is able to jointly address (a) boundary detection (b) saliency detection (c) normal estimation (d) semantic segmentation (e) human part segmentation (f) human boundary detection (g) region proposal generation and object detection in 0.7 seconds per frame, with a level of performance that is comparable to the current state-of-the-art on these tasks.

Bio:
Iasonas Kokkinos obtained the Diploma of Engineering in 2001 and the Ph.D. Degree in 2006 from the School of Electrical and Computer Engineering of the National Technical University of Athens in Greece, and the Habilitation Degree in 2013 from Université Paris-Est.

In 2006 he joined the University of California at Los Angeles as a postdoctoral scholar, and in 2008 joined as faculty the Department of Applied Mathematics of Ecole Centrale Paris (CentraleSupelec). He is currently an associate professor in the Center for Visual Computing of CentraleSupelec and is also affiliated with INRIA-Saclay in Paris. His research activity is currently focused on deep learning and efficient algorithms for object detection.

He has been awarded a young researcher grant by the French National Research Agency, serves regularly as a reviewer for all major computer vision conferences and journals, and is an associate editor for the Image and Vision Computing and the Computer Vision and Image Understanding journals.