A Multi-departmental Research Network for Brain Research
Author: V. Srinivasa Chakravarthy
CNSLab wishes all the readers of NeuroTales, Happy New Year 2025 🙌🥳
This is an informal proposal that aims at creating a virtual Network of researchers from the Institute, who will collaborate to create a comprehensive, multidisciplinary Brain Theory. These researchers can come from many departments. And this document may be treated as an open invitation to join the Network.
After going through the document, if you are interested in joining the Network, please write to me directly (schakra@ee.iitm.ac.in). I will make appointments with you and, over cup after thoughtful cup, - or in online mode if that’s not possible, - we can discuss collaboration.
Brain Modeling is open for all!
There are two ways to look at brain modeling. One way, the traditional way, sees it as something done in the biotechnology department, as a part of computational biology stream etc.
The other approach views the brain as a complex object that invites scrutiny by faculty from nearly every department. Some may study it as basically an interesting problem in biology while others may come to regard it as a template or a metaphor for problems in their native departments.
A complex multi-faceted object like the brain poses challenges that cut across a wide array of engineering disciplines. It is a computing system, an aspect constantly celebrated in popular media; it is an electrical system with all the electrical activity of neurons; it is a chemical system with all the intra- and intercellular chemical signaling; it is a thermal system with its complex heat removal machinery that accompanies any large computing machine; it is a miracle of construction engineering with its intricately generated and maintained microstructure; it is a mechanical system that provides ideal mechanical conditions for the results of learning to take a physical shape; it is a wonderful material system since it needs to be optimally poised between “not too stiff, not too runny,” a material answer, so to speak, of the “stability-plasticity dilemma.”
As far as engineers are concerned, there is something in the brain for everybody. Let’s see exactly how.
AI and Computer Science
If you are looking for connections between brain modeling and departments, AI and CS would be the natural first choice. Although Deep neural networks (AI is full of them) were originally inspired by the brain, beyond the general outlines, the AI experts didn't bother to take much from the brain in the design of AI systems. However, there is a recent call to look back to the brain for inspiration.
The problem is, AI experts often know little about neurobiology, a lacuna reciprocated by neurobiologists.
Importantly, two things may be borrowed from brain science into AI:
Creation of appropriately abstracted whole brain models can be a great source of inspiration for, and innovation in, AI systems. Such whole brain models can drive autonomous vehicles/ships/submersibles/drones etc.
1. Architecture: why do mammalian brains have certain standard components like hippocampus, cerebellum, cortex etc?
Is it possible to abstract them and use them in AI systems?
2. Dynamics: brain dynamics, of all sorts of species, is often resolved into various frequency bands (alpha, beta, gamma etc), which are given special functional roles. There is no equivalent to it in AI systems.
Is there some compelling need to create AI systems with brain-like oscillatory dynamics?
The above line of work will lead to a search for a brain-inspired neural network that will serve as a Universal Agent, a generic agent that can be trained or evolved to perform a wide variety of tasks. A kind of a von Neumann architecture of brain models. It will be a fundamental research problem in both neuroscience and computer science.
Malekmohamadi Faradonbe, S., Safi-Esfahani, F., & Karimian-Kelishadrokhi, M. (2020). A review on neural turing machine (NTM). SN Computer Science, 1(6), 333.
Electrical Engineering
Neuromorphic circuits and VLSI:
People have been working on hardware realizations of spiking neural networks (SNN) for a long time. SNN chips are also available. However, there is a recent surge of interest in computing with oscillating circuits (Csaba and Porod 2020). Apparently John von Neumann had suggested in the ‘50s, that coupled oscillator circuits can be used to represent binary information and even patented the idea. Had that philosophy been pursued, we would now have more brain-like computer hardware. But it is never too late.
Csaba, G., & Porod, W. (2020). Coupled oscillators for computing: A review and perspective. Applied physics reviews, 7(1).
There is a whole field of neuroscience known as Neuromodulation that refers to electrical stimulation of the brain, at various depths (deep in the brain, on the cortex, on the scalp etc). The aim of such stimulation is to control some bio parameter like for example rigidity in a joint, or depression. Bold experiments are being attempted. (I once met this chap who has a startup right outside UC Berkeley campus - he was trying to stimulate the brain using a large number of electrodes and download the state of one brain into another.) This is often done in classical biology style i.e. empirically by trial and error. Makes sense to try and do this using models. This is a great problem to tackle for control theorists - particularly of the variety who deal with control of high-dimensional, nonlinear, network systems.
Why do people always describe the brain as a circuit? For many it isnt. It is an electrical medium in which electrical waves slosh back and forth with certain characteristic wavelengths and frequencies . Modeling electrical wave propagation in the brain to understand brain waves is a huge and exciting field. Those who like to play with Maxwell equations are exactly the players that this topic needs. This line of work also connects to the hot topic of EM field theories of consciousness.
Cabral, J., Fernandes, F. F., & Shemesh, N. (2023). Intrinsic macroscale oscillatory modes driving long range functional connectivity in female rat brains detected by ultrafast fMRI. Nature Communications, 14(1), 375.
Nunez, P. L., & Srinivasan, R. (2006). Electric fields of the brain: the neurophysics of EEG. Oxford University Press, USA.
Signal and Image Processing:
There are a large number of well-known applications of signal and image processing in brain modeling. If we include functional imaging techniques like fMRI, MagnetoEncephalogram (MEG) etc, then we can include video processing also. A sample problem - how do we determine that a patient is adequately anaesthetised on the surgical table purely based on EEG recordings? These methodologies have a great application potential.
Sanei, S., & Chambers, J. A. (2021). EEG signal processing and machine learning. John Wiley & Sons.
Communication theory:
A simplistic way of looking at the brain is as a sensory (input) - motor (output) system. All sorts of sensory streams, each with its characteristic frequency bands, flow into the brain - e.g. visual (5 Hz- 35Hz) and auditory (20 Hz - 20,000 Hz). The brain itself operates in a fairly narrow band (0.1 Hz to 500 Hz) compared to the artificial signal processing world. These signals are modulated into the native frequencies of the brain, propagated through a complex network of channels, with great variability in operating frequencies, and finally demodulated to produce motor movements. This kind of a perspective of brain function opens the doors to an extensive application of concepts from (analog) communication theory. Only some vague beginnings have been made (Akam and Kullman, 2014) and the field is wide open.
Akam, T., & Kullmann, D. M. (2014). Oscillatory multiplexing of population codes for selective communication in the mammalian brain. Nature Reviews Neuroscience, 15(2), 111-122.
Mechanical Engineering
Continuum Mechanics:
Both simple neural network models and detailed biophysical models of learning and memory in the brain, describe learning as a matter of synaptic plasticity. However, we often forget that this functional plasticity of the synapse is accompanied and often preceded by structural plasticity. This structural plasticity is mediated by axons snaking through the neural tissue to find new structures with which to connect. For this to happen, the neural tissue must be easily “diggable” from the axon’s point of view, pointing to a line of work that lays out a long red carpet to mechanical engineers.
Over a half a century ago, Ralph Landauer argued that in computing devices (irrespective of their physical realization) that perform irreversible computations, there must be an inevitable loss of energy as heat, a suggestion that instantly created a bridge between the brain science and thermodynamics. It is also known, for a long time now, that the brain responds to sensory stimuli, not only in terms of electrical neural responses and of blood volume changes, but also in terms of a local, transient thermal responses. An obvious invitation to thermodynamicists and thermal engineers.
Landauer, R. (1961). Irreversibility and heat generation in the computing process. IBM journal of research and development, 5(3), 183-191.
Ocean and Aeronautical Engineering
It would be a stretch indeed, to link these fields to brain science, but not impossible. The connection emerges from fluid mechanics, a core topic in the above two branches of engineering. 17th century European thinkers, like Descartes for example, had a strange way of explaining reflex action. If you stick your finger in fire, for example, the signal triggers flow of animal spirits (read fluids) from the hand to the brain, where more fluids are displaced, ultimately precipitating a return flux of fluid that deflects the hand away from the fire. This early obsession with fluids to explain human motor action perhaps has its roots in hydraulic engineering, which was the only method of motorised action that Europeans knew at that time. Of course all that was summarily (and hastily?) trashed after the advent of an electrochemical theory of neural function.
Descartes explains how bodily motion occurs. (https://thereitis.org/descartes-discourse-on-method/descartes-discourse-jpg/)
But maybe the “fluid theory" still has some juice left in it? There is evidence from neurobiology that electrical stimulation of neural tissue is accompanied by redistribution of water, mediated by swelling of a special type of cells called astrocytes (Sykova 1997). Interestingly, - though it is a bit of a stretch, - the seminal experiments by Jagadish Chandra Bose in the area of plant electrophysiology, nearly a century ago, on mimosa pudica (touch-me-not plant) showed that, the link between electrical signaling and motor movements of the leaves, is mediated by a slight increase in turgidity of the leaf. Considering the above, is it possible that non-neural cellular structures in the brain implement some kind of a liquid computer, which is perhaps evolutionarily older than the more recent electrochemical one? A challenge open to fluid mechanics experts.
Bose, Jagadish Chandra, Motor mechanisms of plants, 1928.
Syková, E. (1997). The extracellular space in the CNS: its regulation, volume and geometry in normal and pathological neuronal function. The Neuroscientist, 3(1), 28-41.
Civil Engineering
Construction Engineering
We can think of the brain as an extreme example, a template, of a complex constructed object. Other organs are more or less blobs of cells. Of course one cannot downplay the value of certain structures like the air sacs in the lung, or the nephrons in a kidney with their interesting physiological functions. But the brain is in a different league. How does the axon of a certain type of a ganglion cell in the retina know how to cross vast distances (of entire centimeters!) and make connections with a particular type of a cell, in a particular layer (1 of 6), of a particular region (1 of 50) in this structure called thalamus? That’s construction at genius level! Construction engineers may want to approach this god-like object with utmost reverence.
Architecture
There is a lot of work done on how architecture affects the brain.
One may also think of the brain itself as an architecture and compare it with a built up space. I pursued a bit of speculation on those lines some time ago in the blog article.
Brain science isn't so parochial that it cares only for engineers. The net may be cast wider, expanding to non-engineering disciplines.
The Physical Sciences
Only an abstract theory of the brain describes it in purely computational terms. But a more realistic one will have to have its roots in Physics, Chemistry, Biology and expressed in the language of Mathematics. Brain’s connection to the physical sciences is obviously too deep and extensive to describe in a few paragraphs. Two particular instances for the sake of concreteness - one from math and the other in physics.
In mathematics, in the broad area of dynamical systems, there is a need to develop theories related to networks of chaotic oscillators, which can be applied to neurobiology.
In physics, stat mech has been a source of important ideas in brain theory from the early days of the connectionist revolution. The recent Nobel given to John Hopfield was one such an instance. Perhaps there is a lot more out there where it all came from. And then there is the eternal question of a quantum theory of the brain which can explain consciousness.
Hopfield, J. J. (1982). Neural networks and physical systems with emergent collective computational abilities. Proceedings of the national academy of sciences, 79(8), 2554-2558.
Satinover, J. (2001). The quantum brain: the search for freedom and the next generation of man.
Humanities and Social Sciences
Linguistics
In a country that speaks over 1600 languages and writes/reads in over a dozen scripts, linguistics touches every aspect of our life - brain science is no exception. A study of how the brain learns languages - first language (L1), second language (L2), and, yes, third language (L3) and beyond - their representations and their interrelationships in the brain, is an area of inexhaustible interest in the Indian context. Difficulty or failure to learn languages, - as in illiteracy or dyslexia - may be researched by combining computational modeling with empirical studies, an exercise that might deliver immense fruits to Indian society. The sons and daughters of Panini may swing into action.
Management Sciences
EconomicsTo find a topic that connects neuroscience and economics is a no brainer - it's called neuroeconomics. The massive world-wide advertisement industry will pay you big bucks if you have a superior method that can predict and even manipulate how people make economic decisions.
To conceive of a deeper topic that links these two fields, one has to regard the brain as a metaphor. The brain is often described as an energy glutton, in comparison to other organs in the body. However, compared to computers of similar stature, the energy it consumes is a mere whiff. Circulation of energy (read, blood) in such a complex network (Pradhan & Chakravarthy 2011) can be viewed as a metaphor for circulation of money in a large society like ours. Large models of neuro-vascular networks may help understand macroeconomic phenomena.
Glimcher, Paul W.; Fehr, Ernst, eds. (2014). Neuroeconomics: decision-making and the brain (Second ed.). Amsterdam: Elsevier Academic Press.
Pradhan, R. K., & Chakravarthy, V. S. (2011). Informational dynamics of vasomotion in microvascular networks: a review. Acta physiologica, 201(2), 193-218.
- The End… at least for now -
Comments
Post a Comment