El DuvelleJun 1, 2023

Come on - does no one care about SFN? 👀

From: @elduvelle
https://neuromatch.social/@elduvelle/110468701179097408

El Duvelle (@[email protected])

#SfN2023 abstract submissions are officially open! More info: https://www.sfn.org/meetings/neuroscience-2023/call-for-abstracts Deadline: Wednesday, June 14, 5 p.m. EDT. Fee: $155 🤔 🤔 + cost of being a member 🤔🙄 (Despite my complaints I’ll probably be there - who else will be?!) Where: Washington DC + poster session online When: November 11-15, 2023. #SfN #Neuroscience

Neuromatch Social
Show threadDan GoodmanJun 1, 2023
@elduvelle never been. Doesn't sound like my cup of tea. Don't plan to go.
Show threadEl DuvelleJun 1, 2023
@neuralreckoning what conferences do you like / go to?
Show threadDan GoodmanJun 1, 2023
@elduvelle smaller ones! Can't cope with that many people. Also these days, ones I can get to by train, trying to avoid flying. Going to CCN in Oxford in August.
Show threadAndrew Glennerster
@neuralreckoning @elduvelle I would love to go to CCN. I live in Oxford, so my carbon footprint would be low.
Jun 2, 2023 at 6:29amWeb
Show threadAndrew GlennersterAug 6, 2023
@neuralreckoning @elduvelle I am interested in computational theories of vision and of brain function as a whole. I have been looking through the #ccn2023 abstracts for examples but not found any yet. Lots of computational modelling, of course, but that is different. Marr’s computational theory has not survived well, but he made a compelling case for one. What are our best attempts today? https://www.cell.com/current-biology/pdf/S0960-9822(02)01204-6.pdf
Show threadAndrew GlennersterAug 13, 2023
@elduvelle_neuro i have been reading @Neurograce’s chapter on grand unified theories of the brain. A GUT is not quite the same as a computational theory since a CT should be a principled approach derived from the constraints on the inputs and outputs to a system. At least in physics, where the term GUT is most often used, the CT approach is not possible. But it is for neuroscience. The ideas in the chapter were quite disparate (a bad sign) and unconvincing.
Show threadAndrew GlennersterAug 26, 2023
@elduvelle_neuro @Neurograce Helen Barron just gave a fascinating talk at CCN2023 describing elegant, compelling and herculean work But it is a good example of the discussion above. The slides in the introduction shown here did not came close to what Marr would describe as a computational theory (in this case of of taking shortcuts).
Show threadEl Duvelle NeuroAug 26, 2023
@ag3dvr @Neurograce hmm I guess I don’t know enough about Marr to know if the cognitive map would qualify as a theory according to his ideas…
Edit: typo that I just noticed
Show threadAndrew GlennersterAug 26, 2023
@elduvelle_neuro @Neurograce Not on its own, no. "Sets out the goal of a process and an outline of how it can be achieved in principle. This includes defining the input, the output and establishing the constraints that will be used in computing one from the other." https://doi.org/10.1016/S0960-9822(02)01204-6 . Here is an example of an implementation that succeeds on navigational tasks and uses grid cells but not a cognitive map: https://doi.org/10.1038/s41586-018-0102-6. The computational theory is quite different.
Show threadEl Duvelle NeuroAug 26, 2023

@ag3dvr @Neurograce hmm what about this one from #BehrensLab :

The Tolman-Eichenbaum Machine: Unifying Space and Relational Memory through Generalization in the Hippocampal Formation?

Show threadAndrew GlennersterAug 26, 2023
@elduvelle_neuro @Neurograce The TEM representation is a graph of states (nodes) joined by actions (edges) with loop closure rewarded/enforced giving it a more metric structure. This makes it a type of policy network (more than a map) https://doi.org/10.1098/rstb.2021.0448 which is good. Experimentally, we find failures of loop closure/consistency in human navigation, so that bit may not be necessary: https://doi.org/10.1371/journal.pone.0033782 https://doi.org/10.1371/journal.pone.0247818. But, yes, TEM is closer to having a computational theory.