First Meetup

There will be two talks about Machine Learning/Data related topics. Free food, free drinks.

Format:

• 2 presentations (each 30 mins + 15 mins discussion)
• Some time for networking + food + drinks (especially after the 2 talks, but also before and in between)

How to get there:

The building is not the easiest to find!

--> U-Bahn: U1 Georg-Brauchle-Ring - go south-east, pass through a kinda industrial area (car dealer is there), see the MTZ building.

--> Tram: Borstei - from there go north and pass the SWM building along the path to the left of it. See the MTZ building ahead of you on the right.

In any case: a working smartphone helps!

Once you enter the MTZ you will see signs leading you to the meeting room on the 1st floor. See you all there.

Presentations:

We have two interesting presentations of the application of Data Analysis / Machine Learning in the field of Security and Privacy.

Adversarial and Robust Learning

Han Xiao - http://home.in.tum.de/~xiaoh/

Tracking those who track

Jan Stepien - http://stepien.cc/~jan/

Presentations Description:

Adversarial and Robust Learning

Recent years have seen successful machine learning and data mining techniques for recommending items, filtering spam etc. Unfortunately, there are evidences showing that adversaries have found a way to deceive these algorithm. Forexample, spammers may add "good" words to a junk mail to cheat a spam filter. Adversaries may even taint data to mislead the training process, e.g. giving unfair rating to their competitor's product. The threat of adversaries pressures us to question current learning algorithms. Are they really robust in the adversarial setting?

The content of this talk is be based on my previous research; it consists of three parts. First, we will discuss the vulnerability of some popular learning algorithms and adversaries strategies. The second part focuses on developing robust learning algorithm with the presence of adversarial noise. Finally, we tackle the challenge of big data and several recent work on large-scale machine learning algorithm will be introduced.

Bio
Han Xiao is a Ph.D. student at Technischen Universität München major in computer science. He has published several papers in AAAI, ECAI, ECIR, PAKDD,etc. His research interests include Bayesian nonparametrics, margin-based methods, sparse methods and probabilistic graphical models.

Tracking those who track

Data analysis is not only about big data and massive clusters. Small
quantities and personal analytics which we can do on our own can also
lead to interesting findings. To back this claim I'm going to talk about
an experiment I've been conducting for a year now. I've been collecting
data about my laptop's HTTP traffic going to various tracking and
advertisements services. Let's see what can we find in it!

Bio

Alumnus of Warsaw University of Technology. Developer at stylefruits
GmbH. When in front of the keyboard he’s most probably juggling with
bits of Clojure and Ruby. When offline, he can be found outdoors; either
preparing for the next marathon or wandering with a way too big backpack
somewhere in the Alpine or Carpathian wilderness.