Dharani Punithan

Sampling bias in estimation of distribution algorithms for genetic programming using prototype trees

Probabilistic models are widely used in evolutionary and related algorithms. In Genetic Programming (GP), the Probabilistic Prototype Tree (PPT) is often used as a model representation. Drift due to sampling bias is a widely recognised problem, and may be serious, particularly in dependent probability models. While this has been closely studied in independent probability models, and more recently in probabilistic dependency models, it has received little attention in systems with strict dependence between probabilistic variables such as arise in PPT representation. Here, we investigate this issue, and present results suggesting that the drift effect in such models may be particularly severe - so severe as to cast doubt on their scalability.We present a preliminary analysis through a factor representation of the joint probability distribution. We suggest future directions for research aiming to overcome this problem.

Daisyworld in Two Dimensional Small-World Networks

Daisyworld was initially proposed as an abstract model of the self-regulation of planetary ecosystems. The original one-point model has also been extended to one- and two-dimensional worlds. The latter are especially interesting, in that they not only demonstrate the emergence of spatially-stabilised homeostasis but also emphasise dynamics of heterogeneity within a system, in which individual locations in the world experience booms and busts, yet the overall behaviour is stabilised as patches of white and black daisies migrate around the world. We extend the model further, to small-world networks, more realistic for social interaction – and even for some forms of ecological interaction – using the Watts-Strogatz (WS) and Newman-Watts (NW) models. We find that spatially-stabilised homeostasis is able to persist in small-world networks. In the WS model, as the rewiring probabilities increase even far beyond normal small-world limits, there is only a small loss of effectiveness. However as the average number of connections increases in the NW model, we see a gradual breakdown of heterogeneity in patch dynamics, leading to less interesting – more homogenised – worlds.

Self-organizing spatio-temporal pattern formation in two-dimensional daisyworld

Watson and Lovelocks daisyworld model [1] was devised to demonstrate how the biota of a world could stabilise it, driving it to a temperature regime that favoured survival of the biota. The subsequent studies have focused on the behaviour of daisyworld in various fields. This study looks at the emergent patterns that arise in 2D daisyworlds at different parameter settings, demonstrating that a wide range of patterns can be observed. Selecting from an immense range of tested parameter settings, we present the emergence of complex patterns, Turing-like structures, cyclic patterns, random patterns and uniform dispersed patterns, corresponding to different kinds of possible worlds. The emergence of such complex behaviours from a simple, abstract model serve to illuminate the complex mosaic of patterns that we observe in real-world biosystems.

Phase transitions in two-dimensional daisyworld with small-world effects- A study of local and long-range couplings

Watson and Lovelocks daisyworld is a coupled biotic-abiotic feedback loop exhibiting interesting planetary ecodynamics. Previous studies have shown fascinating spatio-temporal dynamics in a 2D daisyworld, with the emergence of complex spatial patterns. We introduce small-world effect into such a system. Even a small fraction of long-range couplings destroys the emergent static pattern formation, leading to completely coherent periodic dominance as observed in fully-connected graphs. This change in daisyworld behaviour depends only on the small-world effect, independent of the means by which they are induced (Watts-Strogatz, Newman-Watts and smallest-world models). The transition from static patterns in grid worlds to periodic coexisting dominance in small-worlds is relatively abrupt, exhibiting a critical region of rapid transition. The behaviours in this transition region are a mix of emergent static spatial patterns and large-scale pattern disruption.

Predicting the progression of IgA nephropathy using machine learning methods

We predict the progression of Immunoglobulin A Nephropathy using three of the most widely used supervised classification machine learning algorithms: Classification and Regression Trees, Logistic Regression (in two different forms), and Feed-Forward Neural Networks. The problem is treated as a classification problem, of predicting progression to end-stage renal disease in the ten years following initial diagnosis. All four methods yielded good classifiers, with AUC performance between 0.85 (decision tree) and 0.89 (neural network). The results were generally in-line with expectations, with poor kidney performance on presentation, and evident macroscopic and microscopic damage, all associated with poorer prognosis. However, the association between normal systolic blood pressure status and poor prognosis, for some patients under specific conditions, was entirely unanticipated, and warrants further investigation.

Collective Dynamics and Homeostatic Emergence in Complex Adaptive Ecosystem

We investigate the behaviour of the daisyworld model on an adaptive network, comparing it to previous studies on a fixed topology grid, and a fixed small-world (Newman-Watts (NW)) network. The adaptive networks eventually generate topologies with small-world effect behaving similarly to the NW model – and radically different from the grid world. Under the same parameter settings, static but complex patterns emerge in the grid world. In the NW model, we see the emergence of completely coherent periodic dominance. In the adaptive-topology world, the systems may transit through varied behaviours, but can self-organise to a small-world network structure with similar cyclic behaviour to the NW model.

Evolutionary Dynamics and Ecosystems Feedback in Two Dimensional Daisyworld

We introduce replicator-mutator mechanisms from evolutionary dynamics into a two-dimensional daisyworld model, thereby coupling evolutionary changes with daisyworld’s bidirectional feedback between biota and environment. Daisyworld continues to self-regulate in the presence of these evolutionary forces. The most interesting behaviours, exhibiting a complex and dynamic dance through space and time in species’ abundance, emerges through the introduction of additive spatio-temporal random perturbations in the form of thermal noise. The balance between ecosystem feedback and fluctuations in the ecosystem determines the spatial coexistence of domains of dominance between daisy species and their mutants or adaptants.

Machine Learning Models and Statistical Measures for Predicting the Progression of IgA Nephropathy

We predict the progression of Immunoglobulin A Nephropathy using three classification methods: Classification and Regression Trees, Logistic Regression, and Feed-Forward Artificial Neural Networks. We treat it as a classification problem, of predicting progression to end-stage renal disease in the ten years following initial diagnosis. We compared classifier performance using ROC analysis. All three methods yielded good classifiers, with AUC between 0.85 and 0.95. The results were generally in-line with expectations, with poor kidney performance on presentation, and evident macroscopic and microscopic damage, all associated with poorer prognosis.

An XML format for sharing evolutionary algorithm output and analysis

Analysis of artificial evolutionary systems uses post-processing to extract information from runs. Many effective methods have been developed, but format incompatibilities limit their adoption. We propose a solution combining XML and compression, which imposes modest overhead. We describe the steps to integrate our schema in existing systems and tools, demonstrating a realistic application. We measure the overhead relative to current methods, and discuss the extension of this approach into a community-wide standard representation.