|
@computingnature | |||||
|
Story time! Single neurons in the brain can’t be depended on for reliable information. Here are some neurons from our recent study, recorded twice in response to the same visual stimuli. Different neurons are active at different times!
biorxiv.org/content/10.110… pic.twitter.com/BmAAkxyb3M
|
||||||
|
||||||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
Ask a neuron what angle the corner of your screen makes and it will say 75 degrees right now, 100 degrees in 5 minutes, and some other random number close to 90 every time you ask. pic.twitter.com/5Uq8fb7U9c
|
||
|
||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
That is not how a computational device should work! Imagine if your calculator gave different answers every time... pic.twitter.com/58dqNwlzkh
|
||
|
|
||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
This makes our lives as neuroscientists hard. Single measurements of neurons are not reliable (gray dots), and we need to repeat the measurements many times to average out the noise (black line). pic.twitter.com/N7WXNcuix6
|
||
|
|
||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
Maybe, we thought, the brain uses some kind of averaging over its millions of noisy neurons to get a clean estimate of what it’s looking at. pic.twitter.com/jzTSA9kyK0
|
||
|
|
||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
If that was true, there would be “magical” combinations of neurons, which averaged would give just the right answer.
|
||
|
|
||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
Can we find these “magical” combinations by looking at the brain while it’s looking at our images? We used a microscope to record the activity of ~20,000 neurons simultaneously. Here is all of them from one session in random colors. pic.twitter.com/iaNkGRUZCD
|
||
|
|
||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
We used linear regression to find weights for each neuron that combine their activities into “super-neurons”. pic.twitter.com/Rxepy0oB69
|
||
|
|
||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
These super-neurons were much less noisy than single neurons. In fact, the super-neurons could tell the difference between 45 and 46 degrees on 95% of the test trials. Can you? pic.twitter.com/wuueSaanHH
|
||
|
|
||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
Imagine asking a mouse to distinguish such small differences... Our colleagues in @BenucciLab actually tried! The mouse could only tell apart angle differences of 29 degrees, which was about 100 times worse than the neurons. pic.twitter.com/MeXXVeiWma
|
||
|
|
||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
Even for humans it’s difficult, but I bet you can see the difference if I make the pictures into a movie. pic.twitter.com/S7nfZEpGbP
|
||
|
|
||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
We conclude that mice have a lot of information in their brains, which are 1000x smaller than ours. pic.twitter.com/xVxz0TW5cB
|
||
|
|
||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
They can’t communicate this information to us, but that does not mean they don’t use it, for example as a first step to another computation. pic.twitter.com/9xxfVQ7nPW
|
||
|
|
||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
We hope to find out in the future what these other computations might be.
|
||
|
|
||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
We publicly shared the data and code from this paper if anyone wants to dig further.
data:figshare.com/articles/Recor…
code: github.com/MouseLand/stri…
|
||
|
|
||
|
|
Carsen Stringer
@computingnature
|
22. lip |
|
The End.
|
||
|
|
||