Scientists discover specific neurons involved in making memories

https://www.discovermagazine.com/the-sciences/scientists-discover-specific-neurons-involved-in-making-memories

by lxm • 4 years ago

56 9 4 years ago

I wish we had a way to infer what led to their specificity. We have rediscovered some neuron types in artificial networks using specific architectures, for example grid-cells as a result of recurrent inputs, or complex cells using convolutional layers. But it's always accidental as a result of tuning the models for specific data types.

netizen-936824• 4 years ago

Could you expand on this rediscovery? Are there papers on the topic? Also are you talking about ML NNs or computational neuroscience NN?

ad404b8a372f2b92• 4 years ago

Neuroscience is not my specialty, I only understand the ML side of it. My understanding is that when our machine learning architectures are specially crafted to model specific types of data, we observe similar neurons as those in the brain because they converge to the same optimal solution to handle this data.

For example, taking into account the translational invariance of image data with convolutional layers leads to similar neurons as the V1 cortex. Training recurrent layers on navigation tasks lead to grid-cells similar to those in the brain.

Some references:

https://arxiv.org/abs/1803.07770

https://hal.inria.fr/hal-01389809/document

https://www.biorxiv.org/content/10.1101/296301v1.full

idleproc• 4 years ago

I don't know if this is related, but I used to ruminate a lot when I was younger.

It seems obvious that the mind loves an neat conclusion to events, like a story with a satisfying ending, as opposed to an ambiguous one that leaves you pondering for a while. Or the part of a song that keeps looping (ear worm) until you play it out in full, or get distracted by something else.

As I've got older though, perhaps due to experience (a larger data set) or understanding (passive introspection), I've become able to tolerate ambiguity a lot more, seeing the world as a vast shade of grey rather than black and white.

Or perhaps my brain is simply malfunctioning(?). Replacing 'hard' boundaries with 'soft' ones, so that my life is no longer a series of striking events but rather a pastiche of recurring themes.

This doesn't seem to be a normal progression in any case. I see plenty of people older and more experienced than me blindly stuck in negative thought patterns, or trying to fit their experiences into boxes labelled black or white.

Ulimately, I believe there is some underlying (physical/chemical) cause for our differences, and the results of this study might be part of the puzzle. But just as whole genome sequencing hasn't led us to a full understanding of the human body (yet), I wonder just how far we've got to go before we have a universal understanding of our mental processes.

smm11• 4 years ago

Thank you for this!

webdoodle• 4 years ago

Long-term exposure to repeated cognitive dissonance?

sjtindell• 4 years ago

I think the world we live in today, being civilized as opposed to jungle living, is much more ambiguous. Our brains are not tuned for it.

curious_cat_163• 4 years ago

Bit of a tangent -- I am not sure about how our brains deal with ambiguity vs sharp boundaries, but we can see a neural network change its tune quite a bit by adjusting this single number (i.e. the temperature parameter of Softmax function).

This trick has been used for knowledge distillation (KD) as well. Hinton et. al. wrote this quite approachable paper on the topic: https://arxiv.org/pdf/1503.02531.pdf

Turns out, one _needs_ nuance when one is trying to do KD. I am not sure if nuance and ambiguity as you describe are the same thing. But, even though it does not _feel_ like it, good generalizations over the lifetime of an organism might require dealing with nuance appropriately.

A subject matter expert might have more to say about this. I cannot imagine that this hasn't been researched in humans.

caycep• 4 years ago

Ueli, the senior author on the paper, is a good guy. A lot of interesting pubs/software on his site(s). Good matlab packages if you are analyzing single unit neuron recordings! (Albeit in matlab…)

I think the gist of this is that they have gotten good and differentiating specific neuron behaviors out of the overall mishmosh of noise when one is able to record from human brains during prep for epilepsy or deep brain stimulation surgery. That being said, I think while describing behaviors of individual neurons is slowly increasing understanding of memory cells, the overarching goal of describing a higher level human memory models remains elusive

Current: https://www.cedars-sinai.edu/research/labs/rutishauser.html

Rutishauser.org https://urut.ch/new/serendipity/

Scientists discover specific neurons involved in making memories

Related Stories