Designing a photometric system to best fulfil a set of scientific goals is a complex task, demanding a compromise between conflicting requirements and subject to various constraints. A specific example is the determination of stellar astrophysical parameters (APs) - effective temperature, metallicity etc. - across a wide range of stellar types. I present a novel approach to this problem which makes minimal assumptions about the required filter system. By considering a filter system as a set of free parameters it may be designed by optimizing some figure-of-merit (FoM) with respect to these parameters. In the example considered, the FoM is a measure of how well the filter system can `separate' stars with different APs. This separation is vectorial in nature, in the sense that the local directions of AP variance are preferably mutually orthogonal to avoid AP degeneracy. The optimization is carried out with an evolutionary algorithm, which uses principles of evolutionary biology to search the parameter space. This model, HFD (Heuristic Filter Design), is applied to the design of photometric systems for the Gaia space astrometry mission. The optimized systems show a number of interesting features, not least the persistence of broad, overlapping filters. These HFD systems perform as least as well as other proposed systems for Gaia, although inadequacies remain in all. The principles underlying HFD are quite generic and may be applied to filter design for numerous other projects, such as the search for specific types of objects or photometric redshift determination.

Two major goals in machine learning are the discovery and improvement of solutions to complex problems. In this paper, we argue that complexification, i.e. the incremental elaboration of solutions through adding new structure, achieves both these goals. We demonstrate the power of complexification through the NeuroEvolution of Augmenting Topologies (NEAT) method, which evolves increasingly complex neural network architectures. NEAT is applied to an open-ended coevolutionary robot duel domain where robot controllers compete head to head. Because the robot duel domain supports a wide range of strategies, and because coevolution benefits from an escalating arms race, it serves as a suitable testbed for studying complexification. When compared to the evolution of networks with fixed structure, complexifying evolution discovers significantly more sophisticated strategies. The results suggest that in order to discover and improve complex solutions, evolution, and search in general, should be allowed to complexify as well as optimize.

This paper presents an evolutionary algorithm with a new goal-sequence domination scheme for better decision support in multi-objective optimization. The approach allows the inclusion of advanced hard/soft priority and constraint information on each objective component, and is capable of incorporating multiple specifications with overlapping or non-overlapping objective functions via logical "OR" and "AND" connectives to drive the search towards multiple regions of trade-off. In addition, we propose a dynamic sharing scheme that is simple and adaptively estimated according to the on-line population distribution without needing any a priori parameter setting. Each feature in the proposed algorithm is examined to show its respective contribution, and the performance of the algorithm is compared with other evolutionary optimization methods. It is shown that the proposed algorithm has performed well in the diversity of evolutionary search and uniform distribution of non-dominated individuals along the final trade-offs, without significant computational effort. The algorithm is also applied to the design optimization of a practical servo control system for hard disk drives with a single voice-coil-motor actuator. Results of the evolutionary designed servo control system show a superior closed-loop performance compared to classical PID or RPT approaches.

In this paper, we analyze the decision version of the NK landscape model from the perspective of threshold phenomena and phase transitions under two random distributions, the uniform probability model and the fixed ratio model. For the uniform probability model, we prove that the phase transition is easy in the sense that there is a polynomial algorithm that can solve a random instance of the problem with the probability asymptotic to 1 as the problem size tends to infinity. For the fixed ratio model, we establish several upper bounds for the solubility threshold, and prove that random instances with parameters above these upper bounds can be solved polynomially. This, together with our empirical study for random instances generated below and in the phase transition region, suggests that the phase transition of the fixed ratio model is also easy.

Evolutionary Robotics: A Review

Chapter 7 investigates how learning interest or relevance to them. Chapter 3 describes several techniques and evolution can interact to produce There appears to be no need to read for evolving autonomous robots adaptive individuals.

The field of adaptive robotics involves simulations and real-world implementations of robots that adapt to their environments. In this article, I introduce adaptive environmentics -- the flip side of adaptive robotics -- in which the environment adapts to the robot. To illustrate the approach, I offer three simple experiments in which a genetic algorithm is used to shape an environment for a simulated khepera robot. I then discuss at length the potential of adaptive environmentics, also delineating several possible avenues of future research.

Classifier systems are now viewed disappointing because of their problems such as the rule strength vs rule set performance problem and the credit assignment problem. In order to solve the problems, we have developed a hybrid classifier system: GLS (Generalization Learning System). In designing GLS, we view CSs as model free learning in POMDPs and take a hybrid approach to finding the best generalization, given the total number of rules. GLS uses the policy improvement procedure by Jaakkola et al. for an locally optimal stochastic policy when a set of rule conditions is given. GLS uses GA to search for the best set of rule conditions. 1 INTRODUCTION Classifier systems (CSs) (Holland 1986) have been among the most used in reinforcement learning.

Classifier systems are now viewed disappointing because of their problems such as the rule strength vs rule set performance problem and the credit assignment problem. In order to solve the problems, we have developed a hybrid classifier system: GLS (Generalization Learning System). In designing GLS, we view CSs as model free learning in POMDPs and take a hybrid approach to finding the best generalization, given the total number of rules. GLS uses the policy improvement procedure by Jaakkola et al. for an locally optimal stochastic policy when a set of rule conditions is given. GLS uses GA to search for the best set of rule conditions. 1 INTRODUCTION Classifier systems (CSs) (Holland 1986) have been among the most used in reinforcement learning.