Deep Learning Meetup Autumn 2017

Dear Deep Learning Enthusiasts,

we gonna have another meetup on 10/20/17, 6 PM (doors).

Location: Microsoft Munich, Walter-Gropius-Straße 5, Main Entry

https://secure.meetupstatic.com/photos/event/a/7/c/a/600_465102954.jpeg

Speakers so far:

1. Georg Martius (Max Planck Research Group Leader, MPI for Intelligent Systems, Tübingen)
   Can we extrapolate? Deep learning for equation identification

In classical machine learning, regression is treated as a black box process of identifying a suitable function without attempting to gain insight into the mechanism connecting inputs and outputs.
In the natural sciences, however, finding an interpretable function for a phenomenon is the prime goal as it allows to understand and generalize results. I will present a novel type of function learning network, called equation learner (EQL), that can learn analytical expressions from data. In addition to interpolation it is also able to extrapolate to unseen domains.
It is implemented as an end-to-end differentiable feed-forward network and allows for efficient gradient based training.
Due to sparsity regularization concise interpretable expressions can be obtained. Often the true underlying source expression is identified even for complicated systems such as robotic arm kinematics etc.

2. Leonid Sigal (Senior Research Scientist at Disney Research Pittsburgh and an Adjunct Faculty member at Carnegie Mellon University)
   Talk Topic: On Semantics and Structured Learning

Abundance of inexpensive recording devices and social media platforms have led to an explosion of visual media. Recognition algorithms that help tag, annotate, search and curate this visual data, at different levels of granularity (categorization, detection, linguistic description), are in high demand. Despite significant progress, particularly as a consequence of resurgence of deep learning architectures, a number of important challenges remain. Perhaps most significant are inefficient fully-supervised nature of learning and general lack of ability to encode rich structural dependencies between output variables.

In this talk, I will discuss my work that attempts to address these challenges. First, I will describe a set of models that can leverage explicit (knowledge base) or implicit (linguistic) semantics to improve data efficiency of learning and enable significant scaling of visual recognition models. Second, I will describe non-parametric structured output networks, a new architecture that we propose in our NIPS 2017 paper, that enables one to combine expressive non-parametric graphical models with rich deep neural network observations.

Short CV: Leonid Sigal is a Associate Professor at the University of British Columbia. Prior to this he was a Senior Research Scientist at Disney Research. He completed his Ph.D. at Brown University in 2008; received his B.Sc. degrees in Computer Science and Mathematics from Boston University in 1999, his M.A. from Boston University in 1999, and his M.S. from Brown University in 2003. Leonid's research interests lie in the areas of computer vision, machine learning, and computer graphics. Leonid's research emphasis is on machine learning and statistical approaches for visual recognition, understanding and analytics. He has published more than 70 papers in venues and journals in these fields (including in PAMI, IJCV, CVPR, ICCV, ECCV, NIPS, and ACM SIGGRAPH). He serves on the editorial boards for a number of journals, and on the organizing committees for ICCV and ECCV.

3. Daniel Heinze (Technical Evangelist inside the Commercial Software Engineering Team at Microsoft)
   In his talk Daniel will show the capabilities of the new Azure Machine Learning service and highlights how to create Deep Learning models with it, from data exploration to model creation and publishing.

4. Ralph Hinsche (Business Development Manager Higher Education & Research)
   Talk topic: TBD

TIA to our host Microsoft, which provides us with a large location, and catering, and to TNG and NVIDIA, for additional sponsoring.

Looking forward to see you all again.

Best, Uwe

PS: We are happy to promote a special pricing for the NVIDIA GPU Technology Conference (Munich, 10-12 October 2017) for meetup members. Conference and valuable hands-on DL training sessions.
https://www.gputechconf.eu/Home.aspx
https://www.gputechconf.eu/Register.aspx
Personal Promo-Code RalphHinscheGTCEU17 -> 25% Rebate (everyone)
Academic pricing + Promo-Code RalphHinscheGTCEU17 -> 62.5% Rebate (academia only)
Special Student codes for higher rebates (students only):
Student, 1 day : 75€
Student, 3 days : 165€
Code: STUFGTCEU17

Enter Code on registration to receive rebate