63

I'm developing neural networks comprised of just 3 to 10 layers of virtual neurons and I'm curious to know if there are any insect brains out there with fewer than a thousand neurons?

  • Are there any tiny creatures with small numbers of neurons?
  • Do neuronal maps exist for those simple nervous systems?
J.Todd
  • 1,881
  • 4
  • 19
  • 33
  • 8
    Rotifers have very simple brains with neuron counts in the low hundreds. unfortunately they are not all that well studied. https://academic.oup.com/icb/article/42/3/660/724027/Rotifers-Exquisite-Metazoans1 – John Mar 24 '17 at 15:13
  • 3
    @Viziionary As an aside, note that number of neurons is not always synonymous with lack of complexity; "simpler" nervous systems in terms of numbers of cells often have more complex interactions between neurons: more bi-directional communication, multiple neurotransmitters involved, etc. For example, in most vertebrate neurons, there is a pretty clear distinction between axons (output) and dendrites (input), with a few fun exceptions. However, the distinction between axons and dendrites isn't always straightforward in a worm or insect. Good luck! – Bryan Krause Mar 24 '17 at 15:59
  • @BryanKrause The intricacies and differences between neural mechanisms for different purposes are almost infinite. Which is why I use an evolutionary model where a very extensive list of mutations can happen not only to the neural structure but also the composition and mechanisms involved with each neuron and connection individually. Turns out my code handling mutation is almost as complex and extensive as the rest of the code combined. By the way, what did you mean by "bi-directional communication"? – J.Todd Mar 24 '17 at 16:14
  • Cells that both receive inputs and release neurotransmitter via the same processes. This happens to some extent in vertebrate nervous systems as well, but not with the major neurotransmitter systems. – Bryan Krause Mar 24 '17 at 16:19
  • @BryanKrausen Is it like this? – J.Todd Mar 24 '17 at 16:30
  • @Viziionary No, between neurons; that one is about microglia. This isn't my specific area of expertise (I work with mammalian nervous systems), just some peripheral knowledge so I don't know a great source to cite, or even how well those systems are understood. I think, though I am less certain about this, that some c elegans neurons release different neurotransmitters in different contexts - that might be easier to find information about. – Bryan Krause Mar 24 '17 at 17:10
  • From a computing standpoint, trying to model with few neurons isn't necessarily the best approach. Think how our computers are built from huge amounts of virtually identical transistors to get complex behaviour. Few neurons may very well mean that the interactions between the neurons are much more complex than in a "bigger brained" organism. Think the difference between modelling a thousand identical neurons compared with a hundred neurons that each work differently (as in C. elegans). If you're trying to "evolve" a brain, don't look at "least neurons" - those are heavily optimized. – Luaan Mar 27 '17 at 08:32
  • @Luaan My neural model is able to evolve to beat Super Mario Bros (NES) with only 4 layers, and less than 40 neurons. The neurons each individually mutate to take on different characteristics (action potential thresholds, refactory period thresholds, memory length, default built in memory bias, and of course network structure. Just a hand-full of neurons with relatively simple behavior replication (action potential, memory, polarization, depolarization, refactory period, etc) can accomplish fairly impressive tasks in my experience. – J.Todd Mar 27 '17 at 08:41
  • Oh, nevermind; I was really talking about mimicking real life; there's a lot of shortcuts you can take in a game (I've done a few worthwhile experiments), and a lot of things that would be very complex in a real organism can be very easy on a computer (e.g. some memory patterns can be built using the computer's flat memory instead of simulating in a complex network of neurons). I assumed you wanted a virtual organism with as little "cheating" as possible when asking about the "least neuron count organism", which depends a lot on the complexity of its virtual environment. – Luaan Mar 27 '17 at 08:52
  • @Luaan Ah, no I'm not trying to simulate physically accurate neurons in my project, just functionally accurate (which is in itself a significant challenge) – J.Todd Mar 27 '17 at 09:25

4 Answers4

82

Short answer
As far as I know, a complete neural map (a connectome) is only available for the roundworm C. elegens, a nematode with only 302 neurons (fig. 1).

C.elegansconnectome C. elegans
Fig. 1. C. elegans (left, size: ~1 mm) and connectome of C. elegans (right).
sources: Utrecht University & Farber (2012)

Background
Looking at the least complex of animals will be your best bet and nematodes (roundworms) like Caenorhabditis elegans are definitely a good option. C. elegans has some 300 neurons. Below is a schematic of phyla in Fig.2.

You mention insects; these critters are much more complex than roundworms. The total number of neurons varies with each insect, but for comparison: one of the lesser complex insects like the fruit fly Drosophila already has around 100k neurons, while a regular honey bee has about one million (source: Bio Teaching).

Complexity of the organism is indeed an indicator of the number of neurons to be expected. Sponges, for instance (Fig. 1) have no neurons at all, so the least complex of animals won't help you. the next in line are the Cnidaria (Fig. 2). The Cnidaria include the jelly fish, and for example Hydra vulgaris has 5.6k neurons.

So why do jelly fish feature more neurons? Because size also matters. Hydra vulgaris can grow up 15 mm, while C. elegans grows only up to 1 mm. See the wikipedia page for an informative list of #neurons in a host of species.

A decent neuronal connectivity map (a connectome) only exists for C. elegans (Fig. 1) as far as I know, although other maps (Drosophila (Meinertzhagen, 2016) and human) are underway.

References
- Farber, Sci Am February 2012
- Meinertzhagen, J Neurogenet (2016); 30(2): 62-8

phyla
Fig. 2. Phyla within the kingdom of animalia. source: Southwest Tennessee University College

WYSIWYG
  • 35,564
  • 9
  • 67
  • 154
AliceD
  • 52,402
  • 19
  • 174
  • 238
  • 1
    FYI a full body simulation of the entire connectome of the c elegans - openworm

    https://www.youtube.com/watch?v=SaovWiZJUWY

     – StefanS Mar 24 '17 at 23:00
  • 4
    @StefanSzekeres To be clear, that simulation does not model the connectome or include neural activity, it is only a physics simulation of musculature. Their connectome simulation is not complete yet, as the behavior at the synapses is still not completely understood (although this link is dated it still seems to be the case according to the more up-to-date openworm.org). Not quite there yet, but it's an awesome project worth keeping an eye on. It's only a matter of time. – Jason C Mar 25 '17 at 00:16
25

The organism you are looking for is the nematode C. elegans, which always has the same number of neurons, 302, and has been fully mapped, see WormWeb or you can chase original publications from there. C. elegans is particularly suited for this kind of work because it has a constant number of cells which divide in an entirely predictable order and its neurons forms predictable connections. Larger organisms, such as flies, have a variable number of cells and their neurons do not form precisely predictable connections. The immense amount of knowledge about C. elegans, advanced genetic manipulation techniques, and a transparent body also helps.

I am not aware of any insects with such small brains, even a fruitfly has several orders of magnitude more.

Jack Aidley
  • 6,907
  • 32
  • 48
16

I believe there are types of water snail with 8 distinct neurons in a ganglia, there's a bit of information here: molluscs.at. The cell bodies of the neurons are massive, visible under a standard dissecting microscope, so they were popular among early electrophysiologists. I guess there are probably more neurons around the snail, but it's certainly one of the simplest "brains" around...

Oliver Houston
  • 939
  • 4
  • 9
8

Re: insect brain size

Following article has a good summary — in short insects' nervous systems range from 7400 to 850000 neurons:

http://blogs.discovermagazine.com/notrocketscience/2011/11/30/how-fairy-wasps-cope-with-being-smaller-than-amoebas/

There may be some hope for parasitic insects, e.g. Dicopomorpha echmepterygis where male has neither wings nor eyes, it is not inconceivable for their brains to be simpler.

C_Z_
  • 2,485
  • 15
  • 29
Dima Tisnek
  • 181
  • 4