Past Meetup

20th Deep Learning Meetup in Vienna

This Meetup is past

318 people went

Location image of event venue

Details

Dear Deep Learners,

we have a few exciting announcements after the summer break. First, we open the new season with a special themed evening on reinforcement learning. We have the pleasure to present two highly interesting talks introducing to the theory of reinforcement learning and striving into various fields of applications spiked with lots of examples. Also scroll down to the bottom for our new Wiki, VDLM T-shirts and the poster to this meetup.

Talk 1:
Deep Reinforcement Learning: Learning Like a Baby Rather Than a Copier
Eric Steinberger

This talk, given by Deep Reinforcement Learning researcher Eric Steinberger, is about learning from experience. Deep Reinforcement Learning is a subfield of AI that deals with learning from interactions with a simulator. Famous results of Deep RL include AlphaGo, and more recently OpenAI Five, an AI that crushed ex-pros in 5v5 Dota 2. Behind these results are algorithms that make use of batches of environment interaction (i.e. observation+reward pairs) and try to make the best of it. RL presents many challenges that make supervised learning look like a walk in the park: the data distribution changes as the model changes, we can't use model regularization, training in an imperfect simulation of the real world, and how should the AI explore efficiently to find good strategies? Many of these challenges have good solutions to-date, but Deep RL is still a very active field of research.

Talk 2:
They Grow Up So Fast
Peter Ferenczy

The state of public discourse about robots taking over human jobs is ambivalent at best. There's a lot of fear, uncertainty and doubt regarding the transition towards a radically automated future. But, as Morpheus put it, if we are to be prepared for it, we must first shed our fear of it. Moreover, we must understand the techniques to be in a position to drive this change that ultimately relieves humans of repetitive, boring tasks.
In this talk Budapest based Peter Ferenczy will share his enthusiasm and go in technical depth to illustrate the evolution of Deep Learning techniques to demystify how computers got so successful in competitive games like Go, Chess, Shogi and Dota. We will start with the basics of Reinforcement Learning, take a look at neural network architectures used, understand Transfer Learning and a few extras like how it's possible to democratize a relatively resource intensive training process.
Peter is an avid Go player and the Hungarian translator of the original Dota game, and is running an AI Dojo in Budapest.

Hot Topics:
Rene Donner, Head of Machine Learning & Engineering at Contextflow and Alexander Schindler, Data Scientist at the Austrian Institute of Technology, will report from recent advances in Deep Learning.

Join us for networking & discussions in the break and after the talks.
We are thankful for WKO AUSSENWIRTSCHAFT AUSTRIA to host this meetup, in particular Matthias Grabner for the organization:
https://bit.ly/2LjmuI0

---

Meetup Poster
We have created an announcement poster for this Meetup. Feel free to print it and share it in your office:
https://bit.ly/2MSsM62

VDLM Wiki:
Meetup.com clearly focuses on real-life social interaction and their target platforms are mobile phones. Generally, a good thing, but this makes maintaining a knowledge base for our community complicated. Thus, we decided to move our Wiki to Github, because it's stable and familiar. We've copied all content from the Meetup-pages. Feel free to add or edit. This is a community page, let's make it something useful to us.
https://github.com/vdlm/meetups/wiki

VDLM T-Shirts:
We are excited to announce that our sponsor for the October Meetup, ANEXIA Internetdienstleistungs GmbH, is further eager to provide VDLM branded T-Shirts and invites the community to vote on the final design. The two options look awesome and we are looking forward to your decision - so please vote on:
https://bit.ly/2MypGoq

Looking forward to welcoming you.
Alex, Rene, Tom and Jan