From neurons and spike data to network models

For: Mathematically prepared learners entering computational neuroscience who want a selected route through MIT’s full lecture sequence.

What you will work towards

• Identify assumptions and omissions in a neuron model.
• Relate spike-data summaries to a stated modelling question.
• Describe how response and recurrent-network models address different problems.
• Separate fitting or generating a behaviour from validating a biological explanation.

Known viewing time: 9 hr 41 min across 8 steps; 0 step durations unknown. Lengths are metadata, not evidence of review.

Download the ordered reading list

1. 01

   MIT OpenCourseWare · Source-checked annotation

   1. Overview and Ionic Currents

   MIT OpenCourseWare · Michale Fee · advanced · 1 hr 10 min

   Start with the physical ingredients of neuronal currents before replacing them with a simplified mathematical representation.

   Try this

   List the ingredients in the opening model and one observation each is intended to help explain.

   I have completed this step

   Access evidence checkedOriginal source ↗Watch at source ↗
2. 02

   MIT OpenCourseWare · Source-checked annotation

   2. RC Circuit and Nernst Potential

   MIT OpenCourseWare · Michale Fee · advanced · 1 hr 20 min

   Translate those ingredients into the resistance-and-capacitance representation used by subsequent lectures.

   Try this

   Draw the correspondence between a component of the model and the feature it represents. Mark one simplification explicitly.

   I have completed this step

   Access evidence checkedOriginal source ↗Watch at source ↗
3. 03

   MIT OpenCourseWare · Source-checked annotation

   3. Nernst Potential and Integrate and Fire Models

   MIT OpenCourseWare · Michale Fee · advanced · 1 hr 6 min

   Use the circuit preparation to examine an integrate-and-fire model rather than treating it as an unexplained equation.

   Try this

   State what event counts as a model spike and one biological feature the representation does not attempt to reproduce.

   I have completed this step

   Access evidence checkedOriginal source ↗Watch at source ↗
4. 04

   MIT OpenCourseWare · Source-checked annotation

   7. Synapses

   MIT OpenCourseWare · Michale Fee · advanced · 1 hr 18 min

   Add interactions through synapses before proceeding from single-neuron representations to populations and networks.

   Try this

   Compare two model situations that differ in synaptic input. Predict a difference and state the assumption behind it.

   I have completed this step

   Access evidence checkedOriginal source ↗Watch at source ↗
5. 05

   MIT OpenCourseWare · Source-checked annotation

   8. Spike Trains

   MIT OpenCourseWare · Michale Fee · advanced · 57 min

   Connect the models to the spike signals and summaries available in recorded data.

   Try this

   Specify a data summary and a question it could answer. Describe information that the summary might discard.

   I have completed this step

   Access evidence checkedOriginal source ↗Watch at source ↗
6. 06

   MIT OpenCourseWare · Source-checked annotation

   9. Receptive Fields

   MIT OpenCourseWare · Michale Fee · advanced · 1 hr 18 min

   Now use explicit response models to connect sensory inputs with neuronal activity.

   Try this

   Identify the stimulus and response variables in a receptive-field model. Propose a new stimulus that would test its limits.

   I have completed this step

   Access evidence checkedOriginal source ↗Watch at source ↗
7. 07

   MIT OpenCourseWare · Source-checked annotation

   18. Recurrent Networks

   MIT OpenCourseWare · Michale Fee · advanced · 1 hr 19 min

   Extend from individual responses to recurrent networks that the course relates to memory and decision-making.

   Try this

   Write separately what the network is designed to do and what evidence would be required to treat it as an explanation of a biological system.

   I have completed this step

   Access evidence checkedOriginal source ↗Watch at source ↗
8. 08

   MIT OpenCourseWare · Source-checked annotation

   20. Hopfield Networks

   MIT OpenCourseWare · Michale Fee · advanced · 1 hr 13 min

   Finish with attractor memory models and capacity, turning a network’s limits into an explicit object of study.

   Try this

   State what the model counts as successful storage or retrieval. Identify a failure condition and explain why it is informative rather than merely inconvenient.

   I have completed this step

   Access evidence checkedOriginal source ↗Watch at source ↗

A short starting route

Steps 1, 2, 3. The first three lectures form a coherent physical-to-circuit-to-neuron-model entry before the route branches into data and networks.

Continue more deeply

Return to the full MIT neural-computation course for the intervening conductance, signal-analysis and linear-algebra lectures. Use Neuromatch’s linear-regression tutorial for an inspected Python-based fitting exercise. Continue at source ↗

Limits of this route

The MIT lecture sequence is from 2018. Descriptions, the course’s assignment listing and selected Neuromatch explanatory text were inspected; full videos, all derivations and executable assignments were not. Intermediate course lectures are omitted deliberately, so the listed mathematical prerequisites matter. A model’s successful behaviour does not alone establish biological validity.