Automated Discovery of Relationships, Models, and Principles in Ecology

Abstract

Ecological systems are the quintessential complex systems, involving numerous high-order interactions and non-linear relationships. The most used statistical modeling techniques can hardly accommodate the complexity of ecological patterns and processes. Finding hidden relationships in complex data is now possible using massive computational power, particularly by means of artificial intelligence and machine learning methods. Here we explored the potential of symbolic regression (SR), commonly used in other areas, in the field of ecology. Symbolic regression searches for both the formal structure of equations and the fitting parameters simultaneously, hence providing the required flexibility to characterize complex ecological systems. Although the method here presented is automated, it is part of a collaborative human–machine effort and we demonstrate ways to do it. First, we test the robustness of SR to extreme levels of noise when searching for the species-area relationship. Second, we demonstrate how SR can model species richness and spatial distributions. Third, we illustrate how SR can be used to find general models in ecology, namely new formulas for species richness estimators and the general dynamic model of oceanic island biogeography. We propose that evolving free-form equations purely from data, often without prior human inference or hypotheses, may represent a very powerful tool for ecologists and biogeographers to become aware of hidden relationships and suggest general theoretical models and principles.

A Symbolic Regression Approach to Hepatocellular Carcinoma Diagnosis Using Hypermethylated CpG Islands in Circulating Cell-Free DNA

“…Compared to simple biomarker analysis, machine learning techniques display better performance, which is expected due to their more complex and opaque nature. The symbolic regression approach described in this paper, however, attains similar sensitivity and specificity to conventional ML approaches, is much more transparent and allows for mathematical reasoning of the results in a biological context, which can be a source of useful further research (Narayanan, Cruz Bournazou, Guillén Gosálbez, & Butté, 2022; Cardoso et al, 2020).…”

Towards a cohesive understanding of ecological complexity

“…Approaches typical of complex system science such as Alife, cellular automata, multi-agent models, and genetic programming, based on the idea of interpreting natural processes as computation, remain underrepresented in ecology (21). These approaches have already provided fresh perspectives on traditional dilemmas including the stability-diversity relationship, critical thresholds in habitat loss and fragmentation, the evolution of maladaptive characters, and more (9,21,87).…”

“…While these approaches will remain important for the study of complexity in ecology, there are emergent perspectives that will complement and expand these traditional views. For instance, analysis of networks (66,67) and artificial intelligence (87) have been used increasingly often to accommodate the complexity of ecological systems -at times combining the strengths of more than one of these approaches. Notably, studies of complexity are often developed following a reductionist framework, but progressing in our understanding of complexity will require embracing also novel perspectives developed in complexity science (21,88).…”

The use of machine learning in species threat and conservation analysis

“…This means they can be used to optimise the parameters of a defined model fitting it to observed data and, in the form of genetic programming, they can also evolve the model itself, also known as symbolic regression. The latter is still underdeveloped, although it has a high potential interest, for ecological modelling in particular (Cardoso et al , 2020).…”