Skip to content

Ideas notebook#

Purpose: Arguments between a fictional observer about my work, and my attempts to convince them about my work.
Should practice answering some of these quesstions in recordings.

Practical questions#

  1. What is the size of the imaging plane in the Allen data?

Open questions#

  • What do you think the CLA does? Why are place cells there - could they just be a pass
    through? Maybe mixed selectivity?
  • Is there any difference between mixed selectivity and multisensory neurons?
  • Do you have thoughts on the impact of modelling on the field? E.g Burgess's oscillatory
    interference model of grid cell firing (explain place cell firing and grid cell firing).
  • Do you think that fact that many recordings are performed on a stationary animal has big
    impacts? I mean in some sense we are always on the move, and designed to be on the move.

Questions against myself#

Why don't we use fMRI and functional connnectivity estimates to assess communication between brain regions?#

This technique is complementary to the work we are doing. While the fMRI technique focuses on the BOLD signal in some small mm cubed areas of the brain, here, we propose focusing on this on the level of neurons / neural ensembles, and related LFP measures. As such, it does not contrast with the fMRI method, but should instead be seen as complementary.

How to incorporate the biology into ML#

Great question; need to check the literature! Only real thing I can think is trying to set up the DL layers like real layers, since the brain is evidence of hierarchical processing (TODO check literature on this more). The hierarchical structure is suggested by the sparse connections and the formulation of them.

How does the LFP relate to single units#

Complicated. Large relationship between spiking / field potential / volume conduction. The increase or decrease in field potential influences how easily spiking can happen. As such it is an epiphenomenon of spiking in a sense, but both influence eachother.

Possible questions about thesis#

Further points#

  • Glia: Give off calcium signals which can't be imaged (can use a calcium sensitive dye) - they are about 10 times slower than action potentials.
  • Thoughts of ephatic coupling and communication in the brain outside of action potentials.
  • Perhaps the slow movement of signals in the brain allows for an efficient form of information transfer over only having to fast data transfer (like a data centre style).

This is really quite important for the argument about mapping between animal and human (the cell types).
https://www.youtube.com/watch?v=ylt2ngJdegw

'model free' and the difference between computational neuroscience and using computers to study neuroscience.

  1. Can cell types be detected in CI?
  2. What is the link between CI and spikes (ephys) - there is a paper on this?
  3. What of connectomes and plasticity?