awillats / clinc

closed loop identifiability in neural circuits

👋 🌎 Hello World, I'm Adam Willats 🧠 🔄 🖥️

I research the intersection of systems neuroscience, machine-learning and closed-loop control.
Currently contributing better ways to understand the brain through closed-loop control here: @stanely-rozell

See my academic publications: https://orcid.org/0000-0002-0747-5186 , google scholar page

clinc's Issues

Results - refine figures (figure sketches)

• verify what core takeaway they're aligned with
• make the image concrete or clarify what's missing
• write the takeaway via the caption
  • caption can start overly verbose, then get moved to body of text

Intro - representations & reachability

[communication opportunity / minor]
@adam, prewritten

[ad-hoc, open technical]
[important, but may depend on ongoing work, but lots of dependency]
@Adam/@matt copy paste

Intro - causal methods for network discovery from time-series

Challenges [medium priority for a sketch] - @matthewoshaugnessy

• confound
• indirect correlations
• measures of dependence
  • correlation (linear gaussian) vs info theoretic (TE)
  • bivariate vs multivariate
  • @awillats GC in neuro
    literature recap [mechanical, low risk, slow]

Write gap / context / "description of need"

• review suggestions for this from Chris
• describe intersection of several fields (CL, causal, network neuro)
• describe context of similar papers
• highlight unique aspect of this work
• write final statement

for now this has been folded into abstract / exemplars

Intro - Role of interventions in causal inferences

interventions have been identified as a key ingredient in causal inference
the theory is active, plenty of groundwork, but gap in connecting to practicalities of experiment
gap from the neuroscience side in terms of using the established CI theory to navigate experiment design decisions
@matt [high priority for detailed outline, high priority for early discussion, requires calibration in language used depending on audience]
"stronger interventions facilitate stronger inference"

Intro - interventions in neuro

Write Abstract draft

have an existing draft of the abstract

• matt wrote draft
• review matt's draft - @awillats
• review abstracts from similar papers to assess scope, style - @awillats
• resketch outline, key points @awillats
• add section for expected results @awillats
• fill text for new sections @awillats , @matthewoshaughnessy
• review @awillats @matthewoshaughnessy

Methods - sketch outline quickly

• methods overview paragraph
• update section headings in methods outlines
• tackle individual sections of methods detail

Intro - goal for estimating causal interaction in brain

• @awillats paste existing material
• @matthewoshaughnessy synthesize material

Intro - demo distinguishing a pair of circuits - script

[write a script - turn into a more concrete figure ASAP]
[more concrete draft of diagram - less granular substeps - more annotation ]
[HIGH PRIORTIY - important to Chris, important communication task]
@awillats handoff to @matthewoshaughnessy

Discussion - start limitations / assumptions section

see https://github.com/awillats/clinc/blob/main/sketches_and_notation/discussion/limitations_future_work.md

• [~] limitations
• future work
• [~] breakdown of assumptions
  • especially "does linear v.s. nonlinear matter?"