COMS30017 - Computational Neuroscience

Computational Neuroscience

This is a 3rd year Computer Science unit. We will try to understand how the brain works from a computational point of view.

Staff

• Laurence Aitchison (unit director)
• Conor Houghton

Weekly schedule

• Lectures (weeks 1-7)
  • 9am Tuesdays (FRY BLDG G.10 LT)
  • 3pm Thursdays (CHEM BLDG LT1)
• Problem classes (starting week 2, 5th Oct)
  • 10am-11am Wednesdays (MVB 1.11)

Term schedule

• Weeks 1-7: Lectures
• Weeks 2-: Problem classes. These will be held in MVB 1.11 with TAs + sometimes the lecturer. Please bring any questions on problem sheets, past exams etc. to this session! The idea is to ask TAs questions individually, this isn't a "big" QA session where everyone listens to everyone else's questions!
• Weeks 8-10: Coursework time, for those in CS taking the 20-credit coursework version of the unit. (If you aren't in CS, you're not doing coursework!)
• Week 11-12: Consolidation/revision week for students taking the exam.

Coursework

Please see the folder and the pdf. This is only relevant for those in the Coursework part of the unit (see "Exam vs coursework unit" if you're not sure what part of the course you're on).

Links

• Blackboard page (teaching unit)
• Blackboard page (exam unit)
• Blackboard page (coursework unit)

Exam vs coursework unit.

If you're not a CS student, you're doing the exam version of the unit. If you are a CS student, look on your BB course list.

• If you see: COMS30063_2022_TB-1: Computational Neuroscience (with Coursework) 2022 then you're doing coursework.
• If you see COMS30016_2022_TB-1: Computational Neuroscience 2022 you're on the exam unit.

Please get in touch with your school office (for CS, [email protected]) if you have any questions about this. Conor and I don't touch the enrollment process!

You additionally may have a COMS30017_2022_TB-1: Computational Neuroscience (Teaching Unit) 2022, which is common to everyone.

QA

The course has two primary routes for QA:

• Synchronous QA in the problem classes in MVB 1.11 on Wednesdays at 10-11am.
• Asynchronous QA on Teams, (as we did during COVID). You should already have been added to the group "Welcome to COMS30017: Computational Neuroscience (Teaching Unit) 2022/23 (TB-1, A)". You can download the MS Teams app free here, or use their web app (https://teams.microsoft.com/; Chrome seems least buggy...).

Videos

It seems that we can get access to the lecture videos through BlackBoard. Unfortunately, it may take a couple of days for videos to appear here.

• Go to the teaching unit in BlackBoard (COMS30017_2022_TB-1: Computational Neuroscience (Teaching Unit) 2022)
• Go to Re/Play in the menu on the left

Exam

The exam

• is closed book
• is in person
• is handwritten (unless you have AEAs, please get in touch with the school office to confirm these [email protected]).
• allows calculators (with the usual restrictions that they be "non-programmable"; please email [email protected] for any queries about calculators)

Past exams:

• 20/21 main
• 21/22 main
• 21/22 resit

The format is very slightly different to last year:

• There are 12 compulsory short questions.
• In addition, there are 3 long questions. You pick 2 of the 3 long questions. The long questions is really just a collection of a few short answer questions, but where there can be some links across the questions.
• Previous years have multiple choice questions and short answer questions, while this year we just have short answer questions (remember that the long questions are really just a collection of shorter questions). Importantly, the content + difficulty of the MCQs vs short answer questions should be very similar.

Course content

Week 1: Introduction (LA)

	files
Lecture 1	Why brains; Brain anatomy
Lecture 2	Neuron anatomy; Neural communication; Brain recording
Problem Sheet	Problem Sheet

Week 2: modelling neurons 1 (CH)

	material
Lecture 1	The leaky bucket
related slides	slides
related video	youtube

Week 3: modelling neurons 2 (CH)

	material
Lecture 2	Integrate and fire
related slides	slides
related video	youtube
Lecture 3	Integrate and fire / Hodgkin Huxley

Week 4: Synapses and synaptic plasticity (CH)

	material
Lecture 4	more Hodgkin Huxley / synapses
Lecture 5	more synapses
Worksheet	worksheet
Q5 solution	network.py / network.jl

Week 5: Hippocampus + Hopfield networks (LA)

	files
Lecture 1	The Hippocampus and long term memory; Spatial Navigation
Lecture 2	Pattern Separation; Hopfield Networks; Continuous attractors and navigation
Problem Sheet	Problem Sheet
Answers	Answers

Week 6: Visual system + rate coding (LA)

Laurence managed to get COVID last week, so it'll be video lectures while he is recovering. Hopefully back to usual service next week!

Lecture	video	slides
1. Firing rates and receptive fields	16:51 [Stream link]	[pdf]
2. The visual pathway	15:17 [Stream link]	[pdf]
3. Retina	7:08 [Stream link]	[pdf]
4. V1 and the cortical microcircuit	15:46 [Stream link]	[pdf]
5. Topographic maps and sparse coding	20:43 [Stream link]	[pdf]
Problem sheet	---	[pdf]
Answers	---	[pdf]

Week 7: Cerebellum/basal ganglia, perceptrons, Rescorla-Wagner, blocking (LA)

Lecture	video	slides
1. Supervised learning using the delta rule	12:00 [Stream link]	[pdf]
2. Cerebellar anatomy, function and microstructure	16:27 [Stream link]	[pdf]
3. Classical conditioning	19:32 [Stream link]	[pdf]
4. Temporal difference learning and dopamine	12:17 [Stream link]	[pdf]