,
9 tweets,
4 min read
Read on Twitter
@MatMcGann Good question.
I don’t know the precise answer to that question, but in general, it works like this:
- a region/level can learn to represent two different inputs as the same output, if their “meaning” is the same.
- however, …
1/2
I don’t know the precise answer to that question, but in general, it works like this:
- a region/level can learn to represent two different inputs as the same output, if their “meaning” is the same.
- however, …
1/2
@MatMcGann - …that region can also represent them as two different outputs, if it feels they represent the same “feature/object” in different context.
Which is the mode which is selected is highly dependent on how the following region will use the output of the previous region.
Which is the mode which is selected is highly dependent on how the following region will use the output of the previous region.
@MatMcGann Let me draw you an example, I’ll continue in a few minutes.
@MatMcGann In the drawing below, every rectangle is a brain region. Every square is a neuron. In our brain, neurons are organized in columns.
Each region selects a subset of its columns to be excited, based on which "features" it recognized in the input.
1/N
Each region selects a subset of its columns to be excited, based on which "features" it recognized in the input.
1/N
@MatMcGann 2/ Given a "selected column", the neurons in that column which fire depend on the context only.
Therefore, a given set of column represents a "feature", whereas a given set of neurons represents a "contextualized feature".
Therefore, a given set of column represents a "feature", whereas a given set of neurons represents a "contextualized feature".
@MatMcGann 3/ So, in your previous example, the region might represent steering left as "steering towards the side of the lane in context of the UK" or as "steering towards the left".
Probably, both. Or, two different regions might each represent one of the above. Perhaps,
Probably, both. Or, two different regions might each represent one of the above. Perhaps,
@MatMcGann 4/5 Perhaps, different people would represent it differently. There is no fixed way.
@MatMcGann 5/5
Even conditions such as the Autism Spectrum Disorder influence what one considers as "context" and what as "features". (Simplifying, NeuroTypicals tends to "pull" context into features; ASD, the opposite: they have harder time to integrate context into the features).
Even conditions such as the Autism Spectrum Disorder influence what one considers as "context" and what as "features". (Simplifying, NeuroTypicals tends to "pull" context into features; ASD, the opposite: they have harder time to integrate context into the features).
@MatMcGann (I just clarified the last point here; the examples in the thread, especially later on, might help clarifying the difference between signal and context, from the human brain's point of view )
