Arvutusliku neuroteaduse seminar 2020/21 kevad

Computational Neuroscience Seminar (MTAT.03.292)

Seminars: Thursday 12:15, BBB (log into courses to see link)
Questions:

• Oriol Corcoll (oriol.corcoll.andreu@ut.ee)
• Raul Vicente (raulvicente@gmail.com)
• Lab's website: http://neuro.cs.ut.ee
• Mail list: aine.ati.cnsseminar@lists.ut.ee
• Slack: neurocs.slack.com channel #course-cns-seminar.

About the course

The course is mainly addressed to the Master and Doctorate programs. Students interested in acquiring a perspective in modern neuroscience and computational neuroscience research are welcomed. We will cover advanced and recent papers in neuroscience and deep learning. Advanced research-level discussions are expected to ensure after each paper presentation. Depending on the number of participants each student will read 1 or 2 scientific articles and explain those articles to the others. Note that seminars will be recorded, if you have a problem with this let us know beforehand.

Topic focused seminar

The seminar will focus on a specific topic related to computational neuroscience. This semester we will discuss a recent paper by Schmidhuber's lab: On the Binding Problem in Artificial Neural Networks

The binding problem: The inability of existing neural networks to dynamically and flexibly bind information that is distributed throughout the network. Greff, van Steenkiste & Schmidhuber.

Grading criteria

This is a pass/fail course. During the course, you will collect points¹. To pass you need to collect 70 out of 100 points:

• Presenting an article gives you 50 points multiplied by the score from the listener feedback (example: if listener feedback is 16 out of 20, then you get 50*0.8 = 40 points)
• Attending all seminars gives 20 points (~ 1.42 per seminar²)
• 30 points can be collected from the tests (answering all 4 questions correctly is worth ~2.1 points²)