Research group

Agent Systems and Reinforcement Learning

Neural Architecture Search with Reinforcement Learning

12 March 2019

Neural networks are powerful and versatile models that solve many complex tasks in areas of image analysis, speech recognition and synthesis, etc. The problem of picking the correct architecture is still generally unsolved and is often done manually.

At the seminar, we will discuss two papers that describe methods that utilize reinforcement learning algorithms to to generate descriptions for neural networks. These methods allow generation of neural networks that surpass other state-of-the-art architectures.

Papers: https://arxiv.org/abs/1611.02167 and https://arxiv.org/pdf/1611.01578.pdf

Speaker: Aleksandra Malysheva.

Presentation language: Russian.

Date and Time: March 12th, 20:00-21:30.

Place: Times, room 204.

Videos from previous seminars are available at http://bit.ly/MLJBSeminars

Resources