Distributed associative learning in ecological community networks
Distributed associative learning in ecological community networks
Ecological communities are complex, self-organising systems of interacting species exhibiting important and intricate functions. Yet as evolution by natural selection operates within individual species, and not on community structure as a whole, it is not clear whether or how natural selection organises community-level functions. A long-standing open question thus persists: Are there alternative organising mechanisms that would enable us to understand and predict the complex collective behaviours exhibited by natural communities?
One intriguing possibility is that we might better understand community organisation, not through the principles of evolution (because selection doesn’t operate at the community level), but through the principles of learning, driven by selection acting at lower levels of organisation (i.e. coevolution amongst a community’s component species). Specifically, in this thesis we show conditions where natural selection on ecological interactions is functionally equivalent to a simple type of distributed associative learning, unsupervised learning. This learning rule is well-known in connectionist learning models of cognitive systems, such as neural-networks, where it can produce many non-trivial collective behaviours.
We build from this result and simulate simple Lotka-Volterra models of communities and show conditions where these self-organising processes result in community structure and assembly dynamics that exhibit non-trivial functional properties at the community level. We use these models to demonstrate how community organisation can be conditioned by past experience in the same sense as connectionist learning models habituate to stimuli. We demonstrate the capabilities of this process in the ecological model by showing how the action of individual natural selection can enable communities to i) form a distributed ecological memory of multiple past states; ii) improve their ability to resolve conflicting constraints among species leading to higher community biomass; and iii) learn to solve complex resource-allocation problems equivalent to difficult computational puzzles like Sudoku.
We identify distributed associative learning as a mechanism by which natural selection contributes to community organisation, resulting in a range of adaptive behaviours at the community level. This mechanism is not a result of community-level selection (it cannot be; we prevent group selection effects), and is emergent only from Darwinian processes at lower levels (i.e. individual-level selection).
University of Southampton
Power, Daniel, Alastair
83ab0f52-01b7-418f-9cbc-7a4a6962d0d6
March 2019
Power, Daniel, Alastair
83ab0f52-01b7-418f-9cbc-7a4a6962d0d6
Watson, Richard
ce199dfc-d5d4-4edf-bd7b-f9e224c96c75
Power, Daniel, Alastair
(2019)
Distributed associative learning in ecological community networks.
University of Southampton, Doctoral Thesis, 136pp.
Record type:
Thesis
(Doctoral)
Abstract
Ecological communities are complex, self-organising systems of interacting species exhibiting important and intricate functions. Yet as evolution by natural selection operates within individual species, and not on community structure as a whole, it is not clear whether or how natural selection organises community-level functions. A long-standing open question thus persists: Are there alternative organising mechanisms that would enable us to understand and predict the complex collective behaviours exhibited by natural communities?
One intriguing possibility is that we might better understand community organisation, not through the principles of evolution (because selection doesn’t operate at the community level), but through the principles of learning, driven by selection acting at lower levels of organisation (i.e. coevolution amongst a community’s component species). Specifically, in this thesis we show conditions where natural selection on ecological interactions is functionally equivalent to a simple type of distributed associative learning, unsupervised learning. This learning rule is well-known in connectionist learning models of cognitive systems, such as neural-networks, where it can produce many non-trivial collective behaviours.
We build from this result and simulate simple Lotka-Volterra models of communities and show conditions where these self-organising processes result in community structure and assembly dynamics that exhibit non-trivial functional properties at the community level. We use these models to demonstrate how community organisation can be conditioned by past experience in the same sense as connectionist learning models habituate to stimuli. We demonstrate the capabilities of this process in the ecological model by showing how the action of individual natural selection can enable communities to i) form a distributed ecological memory of multiple past states; ii) improve their ability to resolve conflicting constraints among species leading to higher community biomass; and iii) learn to solve complex resource-allocation problems equivalent to difficult computational puzzles like Sudoku.
We identify distributed associative learning as a mechanism by which natural selection contributes to community organisation, resulting in a range of adaptive behaviours at the community level. This mechanism is not a result of community-level selection (it cannot be; we prevent group selection effects), and is emergent only from Darwinian processes at lower levels (i.e. individual-level selection).
Text
E-Version_DanielPowerThesis20190607
- Version of Record
More information
Published date: March 2019
Identifiers
Local EPrints ID: 433535
URI: http://eprints.soton.ac.uk/id/eprint/433535
PURE UUID: 212b93f7-a5b9-4d67-a397-3ae8f92f58be
Catalogue record
Date deposited: 27 Aug 2019 16:30
Last modified: 16 Mar 2024 03:42
Export record
Contributors
Author:
Daniel, Alastair Power
Thesis advisor:
Richard Watson
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics