Neural computational model GrowthEstimate: A model for studying living resources through digestive efficiency

Area of Science:

• Aquaculture and Fisheries Science
• Computational Biology
• Ecological Modeling

Background:

• Accurate estimation of specific growth rate (SGR) is crucial for understanding aquatic living resources.
• Traditional methods face challenges in natural ecosystems due to uncontrolled variables.
• Existing models may not fully capture the complex interplay of factors influencing growth.

Purpose of the Study:

• Introduce GrowthEstimate, a novel neural computational model for precise SGR estimation.
• Utilize recurrent neural networks (reservoir computing) for advanced growth prediction.
• Provide a tool for individual-level growth assessment in diverse environments.

Main Methods:

• Developed GrowthEstimate using reservoir computing type recurrent neural networks.
• Trained the model on four salmonid datasets, incorporating weight, T/C ratio, and condition factor (CF).
• Validated the model with 15% of each dataset and tested on different species.

Main Results:

• GrowthEstimate demonstrated acceptable SGR output ranges across datasets.
• Model performance showed similarity to real SGR values and ranking in wild populations.
• The model effectively estimates growth in individual aquatic organisms.

Conclusions:

• GrowthEstimate offers a significant advancement for precise and comparable growth estimation in wild populations.
• The model minimizes uncertainty in stock assessment and enhances nutritional ecology insights.
• Future improvements aim for broader species and climate zone applicability, fostering collaborative data collection.