Robin Gras

An individual-based evolving predator-prey ecosystem simulation using a fuzzy cognitive map as the behavior model

We present an individual-based predator-prey model with, for the first time, each agent behavior being modeled by a fuzzy cognitive map (FCM), allowing the evolution of the agent behavior through the epochs of the simulation. The FCM enables the agent to evaluate its environment (e.g., distance to predator or prey, distance to potential breeding partner, distance to food, energy level) and its internal states (e.g., fear, hunger, curiosity), and to choose several possible actions such as evasion, eating, or breeding. The FCM of each individual is unique and is the result of the evolutionary process. The notion of species is also implemented in such a way that species emerge from the evolving population of agents. To our knowledge, our system is the only one that allows the modeling of links between behavior patterns and speciation. The simulation produces a lot of data, including number of individuals, level of energy by individual, choice of action, age of the individuals, and average FCM associated with each species. This study investigates patterns of macroevolutionary processes, such as the emergence of species in a simulated ecosystem, and proposes a general framework for the study of specific ecological problems such as invasive species and species diversity patterns. We present promising results showing coherent behaviors of the whole simulation with the emergence of strong correlation patterns also observed in existing ecosystems.

Speciation with gene flow in a heterogeneous virtual world: can physical obstacles accelerate speciation?

The origin of species remains one of the most controversial and least understood topics in evolution. While it is being widely accepted that complete cessation of gene-flow between populations owing to long-lasting geographical barriers results in a steady, irreversible increase of divergence and eventually speciation, the extent to which various degrees of habitat heterogeneity influences speciation rates is less well understood. Here, we investigate how small, randomly distributed physical obstacles influence the distribution of populations and species, the level of population connectivity (e.g. gene flow), as well as the mode and tempo of speciation in a virtual ecosystem composed of prey and predator species. We adapted an existing individual-based platform, EcoSim, to allow fine tuning of the gene flows level between populations by adding various numbers of obstacles in the world. The platform implements a simple food chain consisting of primary producers, herbivores (prey) and predators. It allows complex intra- and inter-specific interactions, based on individual evolving behavioural models, as well as complex predator–prey dynamics and coevolution in spatially homogenous and heterogeneous worlds. We observed a direct and continuous increase in the speed of evolution (e.g. the rate of speciation) with the increasing number of obstacles in the world. The spatial distribution of species was also more compact in the world with obstacles than in the world without obstacles. Our results suggest that environmental heterogeneity and other factors affecting demographic stochasticity can directly influence speciation and extinction rates.

Regularity analysis of an individual-based ecosystem simulation.

We analyze the results of a large simulation of an evolving ecosystem to evaluate its complexity. In particular, we are interested to know how close to a stochastic or a deterministic behavior our simulation is. Four methods have been used for this analysis: Higuchi fractal dimension, correlation dimension, largest Lyapunov exponent, and P&H method. Besides, we use a surrogate data test to reach a final decision about analysis. As we expect, our results show that there is a deterministic and chaotic behavior in ecosystem simulation.

Species abundance patterns in an ecosystem simulation studied through Fisher’s logseries

We have developed an individual-based evolving predator-prey ecosystem simulation that integrates, for the first time, a complex individual behaviour model, an evolutionary mechanism and a speciation process, at an acceptable computational cost. In this article, we analyse the species abundance patterns observed in the communities generated by our simulation, based on Fishers logseries. We propose a rigorous methodology for testing abundance data against the logseries. We show that our simulation produces coherent results, in terms of relative species abundance, when compared to classical ecological patterns. Some preliminary results are also provided about how our simulation is supporting ecological field results.

K-means clustering as a speciation mechanism within an individual-based evolving predator-prey ecosystem simulation

We present a new method for modeling speciation within a previously created individual-based evolving predator-prey ecosystem simulation. As an alternative to the classical speciation mechanism originally implemented, k-means clustering provides a more realistic method for modeling speciation that, among other things, allows for the recreation of the species tree of life. This discussion introduces the k-means speciation, presents the improvements it provides, and compares the new mechanism versus the traditional method of speciation.

A machine learning approach to investigate the reasons behind species extinction

Abstract Species extinction is one of the most important phenomena in conservation biology. Many factors are involved in the disappearance of species, including stochastic population fluctuations, habitat change, resource depletion, and inbreeding. Due to the complexity of the interactions between these various factors and the lengthy time period required to make empirical observations, studying the phenomenon of species extinction can prove to be very difficult in nature. On the other hand, an investigation of the various features involved in species extinction using individual-based simulation modeling and machine learning techniques can be accomplished in a reasonably short period of time. Thus, the aim of this paper is to investigate multiple factors involved in species extinction using computer simulation modeling. We apply several machine learning techniques to the data generated by EcoSim, a predator–prey ecosystem simulation, in order to select the most prominent features involved in species extinction, along with extracting rules that outline conditions that have the potential to be used for predicting extinction. In particular, we used five feature selection methods resulting in the selection of 25 features followed by a reduction of these to 14 features using correlation analysis. Each of the remaining features was placed in one of three broad categories, viz., genetic, environmental, or demographic. The experimental results suggest that factors such as population fluctuation, reproductive age, and genetic distance are important in the occurrence of species extinction in EcoSim, similar to what is observed in nature. We argue that the study of the behavior of species through Individual-Based Modeling has the potential to give rise to new insights into the central factors involved in extinction for real ecosystems. This approach has the potential to help with the detection of early signals of species extinction that could in turn lead to conservation policies to help prevent extinction.

Rule Extraction from Random Forest: the RF+HC Methods

Random forest (RF) is a tree-based learning method, which exhibits a high ability to generalize on real data sets. Nevertheless, a possible limitation of RF is that it generates a forest consisting of many trees and rules, thus it is viewed as a black box model. In this paper, the RF+HC methods for rule extraction from RF are proposed. Once the RF is built, a hill climbing algorithm is used to search for a rule set such that it reduces the number of rules dramatically, which significantly improves comprehensibility of the underlying model built by RF. The proposed methods are evaluated on eighteen UCI and four microarray data sets. Our experimental results show that the proposed methods outperform one of the state-of-the-art methods in terms of scalability and comprehensibility while preserving the same level of accuracy.

Speciation without Pre-Defined Fitness Functions.

The forces promoting and constraining speciation are often studied in theoretical models because the process is hard to observe, replicate, and manipulate in real organisms. Most models analyzed to date include pre-defined functions influencing fitness, leaving open the question of how speciation might proceed without these built-in determinants. To consider the process of speciation without pre-defined functions, we employ the individual-based ecosystem simulation platform EcoSim. The environment is initially uniform across space, and an evolving behavioural model then determines how prey consume resources and how predators consume prey. Simulations including natural selection (i.e., an evolving behavioural model that influences survival and reproduction) frequently led to strong and distinct phenotypic/genotypic clusters between which hybridization was low. This speciation was the result of divergence between spatially-localized clusters in the behavioural model, an emergent property of evolving ecological interactions. By contrast, simulations without natural selection (i.e., behavioural model turned off) but with spatial isolation (i.e., limited dispersal) produced weaker and overlapping clusters. Simulations without natural selection or spatial isolation (i.e., behaviour model turned off and high dispersal) did not generate clusters. These results confirm the role of natural selection in speciation by showing its importance even in the absence of pre-defined fitness functions.

How efficient are genetic algorithms to solve high epistasis deceptive problems

We present an overview of the properties that are involved in the complexity of global combinatorial optimization problems with a focus on epistasis and deceptiveness. As the complexity of a problem is linked to the exploration operators and algorithm used, we propose at first a bibliography of genetic algorithms. We discuss their efficiency to solve global combinatorial optimization problems following the canonical and the statistical approaches. We propose two strategies to handle such problems. In order to evaluate the capabilities and limitations of each of them, we undertake a comparison on a set of problems with varying levels of epistasis and deceptiveness.

Non-unique oligonucleotide microarray probe selection method based on genetic algorithms

In order to accurately measure the gene expression levels in microarray experiments, it is crucial to design unique, highly specific and sensitive oligonucleotide probes for the identification of biological agents such as genes in a sample. Unique probes are hard to obtain for closely related genes such as the known strains of HIV genes. The non-unique probe selection problem is to select a probe set that is able to uniquely identify targets, in a biological sample, while containing a minimal number of probe. This is a NP-hard problem and this paper contributes the first evolutionary method for finding near minimal non-unique probe sets. When used on benchmark data sets, our approach consistently performed better than three recently published methods. We also obtained results that are at least comparable to those of the current state-of-the-art heuristic.