Latest Robots

Wednesday, 4 January 2023

[New post] Dendrocentric Machine Learning?

Site logo image robertmcgrath posted: " Large scale machine learning models are based on mathematical abstractions of biological neural networks.  The highly successful GPT-3 model basically models the synapses of more than 8 million neurons (with preposterously large matrix calculations). " Robert McGrath's Blog

Dendrocentric Machine Learning?

robertmcgrath

Jan 4

Large scale machine learning models are based on mathematical abstractions of biological neural networks.  The highly successful GPT-3 model basically models the synapses of more than 8 million neurons (with preposterously large matrix calculations).

As a former psychology major, I know that there is more to the brain than synapses.  Neurons have nuclei, membranes, and a trunk (axion) with branches (dendrites).  And lots of chemicals inside and out.  (And who knows what quantum effects we don't know about.)

This makes me wonder what is being left out of the commonly used models, and why.

I'm pretty sure the "why" has a lot to do with practical considerations of contemporary engineering.  The synaptic activity is modelled in discretized vectors that block together groups of synapses in blocks of time.  The propagation of values (which natural neurons do via the other anatomical parts) is represented by big matrix computations.  The output represents a spatial layout of synapse, which encodes the result.

This computational model works pretty well, though power consumption is becoming prohibitive.

An alternative design models the behavior of the axon, which propagates sequences of pulses.  This is represented in a specialized chip, confusingly called a "neuromorphic chip".  (This is "morphic" only for selected aspects of a neuron—there could be a lot of different kinds of "neuromorphic chips".)   The output is a temporal pattern (I htink), which represents the behavior over time of each neuron.

This winter Stanford researcher Kwabena Boahen describes another concept: "dendrocentric" computation [1].  This approach models the behavior of dendrites, which, if I recall correctly connect the axon with one or more synapse.

There is much I don't understand about the mathematical representation here.  Biological dendrites have a complicated array of electrochemical behaviors which, presumably encode natural computations.  If I have it right, the model represents the sequence of electrochmical pulses arriving along the dendrite.

The point of this approach is that there is a lot more information encoded in these pulse trains than in the values of synapses.  This is a critical advantage:  Boahen estimates that this could reduce power consumption orders of magnitude compared to GPT-3 like models [2].


Stepping back, all this makes me wonder exactly what we are doing here.  How can emulating only one part of a complex natural system produce reasonable results?  If three different partial emulations all work, at least to a degree, what does this mean?  And what would happen if we combined them into a more comprehensive model?

Above all, it looks to me like there is a missing theory, a theory of how natural neural nets work.  In particular, I think there should a mathematical theory of "learning networks".  Naturally evolved neural networks will be an optimal sub-case, specialized for embodied bio-nets.  But there will be other optima that work completely different but are very efficient and effective for particular cases.  (And who knows what exotic special cases might fall out of the math—microbial networks, "networks" of stars, a theory of human language, a theory of music.)

Phew.

Anyway, I'm intrigued by this growing field of specialized hardware for machine learning.  I admit that my inner software engineer is offended by the inelegant, brute force success of "deep learning".  Throwing massive amounts of ten year old hardware (and electricity) at a problem may work, but it isn't satisfying.

So, yes, please.  Give me some more weird new chip designs.


  1. Kwabena Boahen, Dendrocentric learning for synthetic intelligence. Nature, 612 (7938):43-50, 2022/12/01 2022. https://doi.org/10.1038/s41586-022-05340-6
  2. Charles Q. Choi, Dendrocentric AI Could Run on Watts, Not Megawatts, in IEEE Spectrum - Artificial Intelligence, December 20, 2022. https://spectrum.ieee.org/dendrocentric-learning
Comment
Like
Tip icon image You can also reply to this email to leave a comment.

Unsubscribe to no longer receive posts from Robert McGrath's Blog.
Change your email settings at manage subscriptions.

Trouble clicking? Copy and paste this URL into your browser:
https://robertmcgrath.wordpress.com/2023/01/04/dendrocentric-machine-learning/

Powered by WordPress.com
Download on the App Store Get it on Google Play
at January 04, 2023
Email ThisBlogThis!Share to XShare to FacebookShare to Pinterest

No comments:

Post a Comment

Newer Post Older Post Home
Subscribe to: Post Comments (Atom)

[New post] Giants

...

  • [New post] Slotxo Internet based Picture slot machine game Fun-based activities That will Play
    Piddle Pops posted: "Discover these days more information on Slot machines performance Gear Financial businesses as well as...
  • [New post] The Glass Hotel and Genre Boundaries
    ...
  • [New post] Giants
    ...
  • https://paxorex.blogspot.com
  • https://acciyo.blogspot.com
  • https://sunbrew.blogspot.com
  • https://readingvox.blogspot.com
  • https://neextdraft.blogspot.com
  • https://udimy.blogspot.com
  • https://arcieve.blogspot.com
  • https://diabetesmail.blogspot.com
  • https://quiltingmail.blogspot.com
  • https://downloadallyouwanttutorials.blogspot.com
  • https://increasingmarketingsystem.blogspot.com
  • https://skysportingnewsnationspinquirer.blogspot.com
  • https://politicnewsbusterinsiderpostreview.blogspot.com
  • https://javascripttrendlist.blogspot.com
  • https://teraqiitatail.blogspot.com
  • https://bigpalacenews.blogspot.com
  • https://executivetowernews.blogspot.com
  • https://magnificentplannews.blogspot.com
  • https://businessinboard.blogspot.com
  • https://patriotsscience.blogspot.com
  • https://allinonequantumleap.blogspot.com
  • https://foodandrecipefusion.blogspot.com
  • https://newsletterforeveryone.blogspot.com
  • https://snacksrobinhood.blogspot.com
  • https://dailynewslettersph.blogspot.com
  • https://rankedrama.blogspot.com
  • https://oschinanet.blogspot.com
  • https://nourich.blogspot.com
  • https://phnewsnet.blogspot.com
  • https://structuresusingc.blogspot.com
  • https://foodubers.blogspot.com
  • https://genuinequality.blogspot.com
  • https://techdigitalmedia.blogspot.com
  • https://entertainmenhubtbiz.blogspot.com
  • https://sportsbookwire.blogspot.com
  • https://societycast.blogspot.com
  • https://lifestylesportsreturn.blogspot.com
  • https://natureimpactfactor.blogspot.com
  • https://artnetworth.blogspot.com
  • https://entrepreneurexamples.blogspot.com
  • https://cryptomarketbase.blogspot.com
  • https://btsbiot.blogspot.com
  • https://sexybinikis.blogspot.com
  • https://foreignexchangecurrency.blogspot.com
  • https://classifiedexample.blogspot.com
  • https://bookboons.blogspot.com
  • https://writingdate.blogspot.com
  • https://wamios.blogspot.com
  • https://justmightdiy.blogspot.com
  • https://playfreeonlinegamesmore.blogspot.com
  • https://healthlinefitnessfirst.blogspot.com
  • https://snaptikvideodownloader.blogspot.com
  • https://pokemonunitepc.blogspot.com
  • https://neverthelesskdrama.blogspot.com
  • https://coolantioniq.blogspot.com
  • https://hackerploit.blogspot.com
  • https://ballbreakdown.blogspot.com
  • https://flixsterio.blogspot.com
  • https://fortnitebattleroyaletrack.blogspot.com
  • https://manilaplus.blogspot.com
  • https://davaoplus.blogspot.com
  • https://tutorialsfiles.blogspot.com
  • https://mondaymorningcookingclub.blogspot.com
  • https://gymnearmee.blogspot.com
  • https://windows26.blogspot.com
  • https://millionaireinvest.blogspot.com
  • https://latestkhmernews.blogspot.com
  • https://latestisraelnews.blogspot.com
  • https://latestaustralianews.blogspot.com
  • https://latestirannews.blogspot.com
  • https://latestjapannews.blogspot.com
  • https://latestsaudinews.blogspot.com
  • https://latestfreecourse.blogspot.com
  • https://ikeafurnitureaccessories.blogspot.com
  • https://makeupandbeautyproduct.blogspot.com
  • https://latestpets.blogspot.com
  • https://topecommerceniches.blogspot.com
  • https://latesttexasnews.blogspot.com
  • https://latestufcgame.blogspot.com
  • https://tipweightlossfast.blogspot.com
  • https://latestcancercure.blogspot.com
  • https://philsys.blogspot.com
  • https://phoramensoba.blogspot.com
  • https://latestcupcakes.blogspot.com
  • https://latestgivex.blogspot.com
  • https://latestlottoresult.blogspot.com
  • https://downloadarchived.blogspot.com
  • https://doesports.blogspot.com

Search This Blog

  • Home

About Me

latest robot
View my complete profile

Report Abuse

Blog Archive

  • October 2023 (1228)
  • September 2023 (1871)
  • August 2023 (1663)
  • July 2023 (1819)
  • June 2023 (1774)
  • May 2023 (1651)
  • April 2023 (1598)
  • March 2023 (1753)
  • February 2023 (1419)
  • January 2023 (1661)
  • December 2022 (1507)
  • November 2022 (1620)
  • October 2022 (1463)
  • September 2022 (1332)
  • August 2022 (1370)
  • July 2022 (1493)
  • June 2022 (1331)
  • May 2022 (1450)
  • April 2022 (1438)
  • March 2022 (1366)
  • February 2022 (958)
  • January 2022 (994)
  • December 2021 (1759)
  • November 2021 (3125)
  • October 2021 (3244)
  • September 2021 (3138)
  • August 2021 (3240)
  • July 2021 (1142)
Powered by Blogger.