BangPypers

The Bangalore Python User Group

Talks - Sep, 2017

For September's session, we continued the “Machine Learning/ Deep Learning ” theme. The venue was Synup in Koramangala, and we had 3 speakers. The first two talks were of 40-60 minutes and the last one was around 20 minutes.

The first talk was by Krishna Kant and his talk was titled “Making Deep Learning Models Robust to Adversarial Attacks”. He spoke about how any dataset can be attacked using whitebox and blackbox methods. Whitebox methods such as FGS, an implementation called DeepFool, BFGS and JSMA and Blackbox methods that would basically be using noisy images or adversarial examples generated using the aforementioned methods. He also spoke briefly about defending models against these attacks - using correct classification by training on adversarial examples and rejection of incorrect results.

YouTube video for the talk -

Adversarial

Papers and links for reference -
http://evademl.org/
https://arxiv.org/pdf/1608.04644.pdf
https://arxiv.org/pdf/1511.04599.pdf
https://arxiv.org/pdf/1707.06728.pdf
https://github.com/uvasrg/FeatureSqueezing

The second talk was by Rahul , titled “Introduction to Reinforcement Learning ”
Desc –To-Do–

YouTube videos for the talk -

RL

The third talk was presented by Shrishty Chandra
–To-Do–

YouTube video for the talk -

HowImportWorks

Hope you enjoyed the talks! See you again next time!

Some pics from the meetup - (To-Upload)

p1
p2

The entire Youtube Playlist for the above mentioned talks are available here