Tutorial Development Notes Dump

[TomDonoghue]

VoytekLab Tutorials Plans

Assignments:

• Julio: 0, 6, 7
• Michael: 1, 2, 4
• Sydney: 3, 5, 8

Plan: each person make a branch
work on their 3 notebooks

First pass: copying stuff from the existing tutorials
cleaning it up a little bit
sketching out where there needs to be new things

Conventions

• All in python (convert to Py3, drop Py2 support)
• Follow good code conventions (ex. don't use pylab)
  • Follow our code SOP from Resources
• Each tutorial starts with a number, that places it in order
• imgs & data files are clearly named and go into imgs/ and data/ folders
• Use html div tags: green is 'core sentence overview', blue is 'link out', red is 'important warning'
• Links are in markdown hyperlinks as LABEL. Use good labels.
  • Always use DOI links when available (always for scientific papers)

Notes

• These materials should be tightly integrated with 'Resources'
  • Think of the Resources as the 'master' organizer
  • Tutorials are stand-alone - won't copy stuff that are otherwise in Resources

Tutorials Purpose

The overall goal is to go from some technical knowledge (Python, DSP), into knowing enough to jump into our tools and approaches, in the area of neuro-electrophysiology, particularly around oscillations & 1/d. Scope: Focused around Electrophysiology Field Recordings

Topics:

• Jupyter & Python catch-up
  • Not teaching these things, but quick mention, link out to resources
• DSP
  • Same: not in scope, but mention, and point to resources
• Frequency representations
• Oscillations
• 1/f
• Basic / standard operations: filtering, phase coherence, etc.
• Event-related ideas

In particular, these are on boarding, to direct out to FOOOF, NEURODSP & ByCycle:

Needed to move to FOOOF :

• Presumes: frequency representations and how to get there, oscillations, and (to some extent 1/f)

Needed to move to NeuroDSP / ByCycle"

• Presumes: have neural time series, and some questions about them. Basic DSP, filtering, etc
  • Time series, filtering, consistency vs. burstiness of oscillations, burst detection, spectral analysis / frequency representations

Notebook Plan

• 0: Jupyter & Python

  • Note: mostly quickest possible overview, with links elsewhere (codeacademy, COGS18, COGS108_

• 1: Electrophysiological Time-Series

  • sampling
  • loading (give you some time series in npy file)
  • rhythms (oscillation & burstiness)
  • referencing

• 4: Event Related Potentials

  • event-related potential

• 2: Bands & Filtering

  • How does a filter work

• 3: Spectral Methods / Estimation

  • frequency representation

• 5: Frequency Properties & Interactions

  • phase coupling
  • phase-amplitude coupling

• 6: Time Series Properties

  • colored noise
  • 1/f
  • note: cover this in time domain & frequency domain

• 7: Noise, Filters, Band (oscillation mis-interpretations)

  • How can filtering go wrong
  • How band-specific analysis goes wrong
    • Go to FOOOF

• 8: How Frequency Representations Go wrong, non-sinuisoidal

  • Go to NeuroDSP / ByCycle

Goal: by the end of this, are ready to move onto FOOOF, ByCycle, NeuroDSP

Optional specific cases / things to add / link to:

• Getting started with EEG / MNE
• Modelling
• LFP

[TomDonoghue]

These are the notes from the meetings / discussions we had(with tutorial team of @sydney-smith, @ArcadeShrimp, @MichaelTrann and me supervising), for how to build the tutorials, what to put into them, etc.

Note: the ordering of notebooks listed is different from currently ordered, as we played around with the best profession. We also considered merging 7 & 8.

[TomDonoghue]

Tutorial hack people (@voytek, @rdgao, @LeoVascelli) - starting by checking out this issue hopefully gives an overview of where we were coming from.

Through development and internal review of the notebooks / materials from each other, there are also already a group of issue listed of open questions / comments - so that is probably a good place to start in terms of reviewing materials, adding content, and making suggestions. If there is something not currently listed, make a new issue.

Please don't open PRs on notebooks - they are a nightmare to merge: we'll assign issues to specific people who can work on the notebook and push, with a lock on anyone else working on them, to avoid merge issues.