I love this too. And yet, it would appear that linear integration plus a single nonlinearity can account for quite a bit of variance in a neuron’s responses: https://www.sciencedirect.com/science/article/pii/S0896627318307372 …
-
-
Replying to @tyrell_turing @GunnarBlohm
So you are telling us our model isn't good because it disagrees with another model?
1 reply 0 retweets 5 likes -
Replying to @KordingLab @GunnarBlohm
Yup.
Nah, I’m not telling you your model isn’t good. But I think you’re asking a different question: you’re asking what *can* neurons do, the paper I linked was asking which abstract models can match more biophysical ones.1 reply 0 retweets 8 likes -
I raise this only because I think it’s important to remember that highly abstract units may actually still be good models for many purposes. That doesn’t in any way invalidate your work here.
1 reply 0 retweets 5 likes -
Replying to @tyrell_turing @GunnarBlohm
I think there may be a major problem in the literature. If channel densities are learned, which I believe the are, then simulating the average channel density may project a very misleading image. Imagine neural networks where all weights are the same and not learned...
1 reply 1 retweet 5 likes -
Now why do I believe they are learned? Dl theories assume that dA/dW is calculated. But once you calculate dA/dW it is very easy to calculate dA/dC where C is some channel density or property. It is essentially the same computation.
2 replies 0 retweets 3 likes -
Replying to @KordingLab @GunnarBlohm
I agree. I’d be shocked if channel densities aren’t learned.
1 reply 0 retweets 3 likes -
But, would that necessarily render the result I linked invalid?
1 reply 0 retweets 4 likes -
Are you suggesting that with more variation in channel density between branches a linear-> nl model wouldn’t match a biophysical model well?
1 reply 0 retweets 5 likes -
Replying to @tyrell_turing @GunnarBlohm
Yes! Learning would enforce (strong) nonlinearity where useful.
1 reply 0 retweets 5 likes
-
-
On the plus side
#jobSecurityForAdam0 replies 0 retweets 3 likesThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.