Monday, November 29, 2021

Signal to Noise for the Win


A new model for how the brain perceives unique odors
Oct 2021, phys.org

So here is a scientist, a physicist by name, with an interest in the information-processing abilities of biological, and neurological systems. It didn't take him long, I would imagine, to realize that olfaction is a prime model for complex information-processing systems; it was the first sense, used by the first bacteria to detect chemicals in the primordial soup, later used by animals to make a big, complex mammal brain. If you want to look at how information processing happens in biological systems, this would be the ideal place to look.

But first, you have to throw into the garbage everything you know about olfaction already, which should be easy for a computational information scientist. What they did here was to look far out, all the way out -- beyond molecules and their myriad physico-chemical characteristics, of which the molecules themselves number in the billions; beyond genetics and their 30% variation across the global population; beyond cultural effects that are almost too surreal to quantify for methodical research purposes, like where Americans prefer peppermint since it's associated with candy, yet older folks from England don't like mint since it's associated with pain-relief products of their era, or the easier comparison of preference for durian fruit or √Čpoisses de Bourgogne cheese.

Too messy, said the computational information scientist. And so they put it all in the blender, all those variables together. And they called it noise. Boil it all down, cancel it all out, all that dirty data of molecules and genes and cultural and personal association. Throw it all in the same bin, and call it noise. That's what they did.

Actually, they didn't call it noise, they called it "context" --

 "If you experience odors in a similar context, even if they were initially rather different in the responses they evoked in the nose, they begin to be represented by similar neural responses so they become the same in your head," Balasubramanian says.

The researchers found that their simplified model could be used to reproduce the same types of results seen in olfaction experiments. It's something that Balasubramanian did not expect to see, as he thought that such a complex process would require "learning and plasticity" in order to adapt and change neural synapses to modify the brain's representation of smells. "We may have found a general strategy of using certain kinds of randomized signals to entrain those effects," he says about their results. "It doesn't have to be just olfaction; it can be elsewhere, too." -medicalexpress

Did you see that? "It doesn't have to be olfaction; it can be elsewhere too." Olfactively-piqued, computational information neuroscientist, where have you been? Proving that the nose-brain is the neural model par excellence, while showing us how it actually works, both at the same time.

*If you want to know more about why olfaction is the ideal model for growing an artificial brain from scratch, it's a constant theme in Hidden Scents.

via University of Pennsylvania: Gaia Tavoni et al, Cortical feedback and gating in odor discrimination and generalization, PLOS Computational Biology (2021). DOI: 10.1371/journal.pcbi.1009479

Post Script:
This study shouldn't be mentioned without this other study, where they taught an artificial network how to smell, via Massachusetts Institute of Technology's McGovern Institute for Brain Research: Peter Y. Wang et al, Evolving the olfactory system with machine learning, Neuron (2021). DOI: 10.1016/j.neuron.2021.09.010. http://dx.doi.org/10.1016/j.neuron.2021.09.010

Post Post Script:
Might as well put this here, since it has to go somewhere -- what happens when you take an information scientist and give them an olfactory science problem? This is what happens. They come up with an answer that is so simple it just makes you look stupid. This is an example, although not a true example, of the other half of the coming dark ages. After a global pandemic, it's inevitable to experience a kind of dark ages, where lots of people died, but way more people got sick, and also a lot of people retired. That's a post-pandemic-pandemic of institutional knowledge loss, like a collective long covid brain fog on our culture -- we forget how to do stuff, because the guy who did it for the past 30 years isn't doing it anymore. That guy either died, got too sick to work, or retired for a million other reasons, of which many of them could be pandemic-related. And there's lots of those guys (and even more of them gals). Who knows what that will look like for us today or tomorrow, but it's happening as we speak, and years from now we might notice, we might even call it the great forgetting. The flip side to the dark ages is the renaissance, which comes from all the new people in new roles and at new jobs. These people are coming in new, with nobody around to teach them how to do things "right," and although that makes for a bumpy road ahead, it also gets you things like this discovery, one of the hardest problems of olfaction taken on by someone who has not much at all to do with olfaction (although he should, because olfaction has been an information science problem all along).

Wednesday, November 24, 2021

Olfaction In Silico


Artificial networks learn to smell like the brain
Nov 2021, phys.org

No kidding, the sensory apparatus that resembles a deep learning neural network can be simulated with a deep learning neural network -- "Artificial networks trained to classify odor identity recapitulate the connectivity inherent in the olfactory system."

The part of our brain that smells is also the most primitive. Before brains were a thing, bacteria performed chemosensory calculations on the primordial soup. As the soup became more complex, so did sensory equipment. Chemosensitive receptors on the surface of a bacterium became antennae, and then became noses, and those noses became seeing, hearing, even speaking brains. But the first version is the one used for smelling. So it shouldn't be a surprise that the first place we see a direct link between the mammalian brain and our artificial instantiation is via olfaction. Nonetheless, the scientists were "surprised to see it replicate biology's strategy so faithfully."

"By showing that we can match the architecture very precisely, I think that gives more confidence that these neural networks can continue to be useful tools for modeling the brain," says Robert Yang, assistant professor in MIT's departments of Brain and Cognitive Sciences and Electrical Engineering and Computer Science [and who collaborated on this project with Columbia neuroscientists Richard Axel and Larry Abbott, btw].

They use an antennae model, and the indispensable fruit fly, but it's all close enough to humans, I mean that's why we use the fruit fly in the first place. They started with some artificial neurons, of the same amount found in a fruit fly. They programmed the neurons to identify odors, and to assign valence (pleasant or unpleasant) to odors. They didn't give the neurons any structure, no information about how to talk to each other, no blueprint on how to process information. Just pre-programmed neurons, thrown into a simulated universe of chemosensation.

In minutes, and in silico, a network emerged to look just like the nose-brain of a fruit fly. All on its own, "an initially homogeneous population of neurons segregated into two populations with distinct input and output connections, resembling learned and innate pathways." 

In other words, it learned how to smell. It took evolution some billions of years to get the fruit fly olfactory system just right. The artificial network did it in minutes. Extrapolations from the study abstract: "This implies that convergent evolution reflects an underlying logic rather than shared developmental principles."


*Their structure used expansion and compression layers, similar to the pyramidal-structure of the nose-brain, and the number of inter-connections was also the exact same number as in the fruit fly, and it worked with both feedforward and recurrent network models, and the networks are plastic, meaning they can learn new odor associations over time.  

via Massachusetts Institute of Technology's McGovern Institute for Brain Research: Peter Y. Wang et al, Evolving the olfactory system with machine learning, Neuron (2021). DOI: 10.1016/j.neuron.2021.09.010

Post Script:
The book Hidden Scents talks about how olfaction is an ideal model for understanding an artificial brain, for an artificial human. 

Although artificial neural networks resemble natural neural activity patterns, like those used by the visual cortex for object recognition, we still don't understand how the visual cortex, or most mammalian neural circuits, are inter-connected. This time, we see how they connect, a connectome of the olfactory cortex.

Post Post Script:
And this study should be kept alongside this other one, where a physics-based computational neuroscientist came up with a pretty simple way to mimic the olfactory cortex, via University of Pennsylvania: Gaia Tavoni et al, Cortical feedback and gating in odor discrimination and generalization, PLOS Computational Biology (2021). DOI: 10.1371/journal.pcbi.1009479