A group of Kyoto University, B.Creation Co., Ltd., Tohoku University, and the University of Tokyo conducted experiments using ecosystem simulations and microbial culture systems. I found evidence that I could use that ability.
In recent years, data analysis methods (artificial intelligence) using neural networks have been actively developed and used in various fields, and their computing power has been evaluated and used.However, ecosystem networks (e.g., interspecies relationships such as eating and being eaten) cannot strictly manipulate interspecific interactions and the number of species, making it virtually impossible to optimize the network according to the purpose of data analysis. Therefore, its computing power and availability were unknown.
The research group applied a technology called reservoir computing, a type of neural network that does not require network optimization, and presented a framework for quantifying the computing power of ecosystem networks.Furthermore, experiments using a culture system using the eukaryotic microorganism Tetrahymena revealed that the population dynamics of Tetrahymena provided conditions for reservoir computing.
Therefore, we input the time-series data of fish population fluctuations obtained from the field into the Tetrahymena population as the temperature change of the culture medium, and tested the prediction of fish population fluctuations in the near future.As a result, Tetrahymena populations can be predicted in the near future with higher accuracy than simple data analysis methods such as linear regression.
This simulation suggests that the greater the number of species, the higher the computational power, indicating the possibility that high biodiversity and high computational power correspond. "The computing power of ecosystems" is expected to shed light on new values of biodiversity.
Paper information:[Royal Society Open Science] Computational capability of ecological dynamics