Catálogo de publicaciones - libros

We present a new multiobjective evolutionary algorithm (MOEA), called fast Pareto genetic algorithm (FastPGA). FastPGA uses a new fitness assignment and ranking strategy for the simultaneous optimization of multiple objectives where each solution evaluation is computationally- and/or financially-expensive. This is often the case when there are time or resource constraints involved in finding a solution. A population regulation operator is introduced to dynamically adapt the population size as needed up to a user-specified maximum population size. Computational results for a number of well-known test problems indicate that FastPGA is a promising approach. FastPGA outperforms the improved nondominated sorting genetic algorithm (NSGA-II) within a relatively small number of solution evaluations.

We applied genetic programming with a lambda abstraction module mechanism to learn technical trading rules based on S&P 500 index from 1982 to 2002. The results show strong evidence of excess returns over buy-and-hold after transaction cost. The discovered trading rules can be interpreted easily; each rule uses a combination of one to four widely used technical indicators to make trading decisions. The consensus among these trading rules is high. For the majority of the testing period, 80% of the trading rules give the same decision. These rules also give high transaction frequency. Regardless of the stock market climate, they are able to identify opportunities to make profitable trades and out-perform buy-and-hold.

The chapter summarizes the practical experience of integrating genetic programming and statistical modeling at The Dow Chemical Company. A unique methodology for using Genetic Programming in statistical modeling of designed and undesigned data is described and illustrated with successful industrial applications. As a result of the synergistic efforts, the building technique has been improved and the model development cost and time can be significantly reduced. In case of designed data Genetic Programming reduced costs by suggesting transformations as an alternative to doing additional experimentation. In case of undesigned data Genetic Programming was instrumental in reducing the model building costs by providing alternative models for consideration.

This chapter derives a population sizing relationship for genetic programming (GP). Following the population-sizing derivation for genetic algorithms in (Goldberg et al., 1992), it considers building block decision-making as a key facet. The analysis yields a GP-unique relationship because it has to account for bloat and for the fact that GP solutions often use subsolutions multiple times. The population-sizing relationship depends upon tree size, solution complexity, problem difficulty and building block expression probability. The relationship is used to analyze and empirically investigate population sizing for three model GP problems named ORDER, ON-OFF and LOUD. These problems exhibit bloat to differing extents and differ in whether their solutions require the use of a building block multiple times.

This chapter presents a tiered view of the roles of structure in genetic programming. This view can be used to frame theory on how some problems are more difficult than others for genetic programming to solve. This chapter subsequently summarizes my group’s current theoretical work at the University of Michigan and extends the implications of that work to real-world problem solving.

This is a narrative describing the implementation of a genetic programming technique for stock picking in a quantitatively driven, risk-controlled, US equity portfolio. It describes, in general, the problems that the authors faced in their portfolio context when using genetic programming techniques and in gaining acceptance of the technique by a skeptical audience. We discuss in some detail the construction of the fitness function, the genetic programming system’s parameterization (including data selection and internal function choice), and the interpretation and modification of the generated programs for eventual implementation.

This chapter describes the notion of Run Transferable Libraries(RTLs), libraries of functions which evolve from run to run. RTLs have much in common with standard programming libraries as they provide a suite of functions that can not only be used across several runs on a particular problem, but also to aid in the scaling of a system to more difficult instances of a problem. This is achieved by a library on a relatively simple instance of a problem before applying it to the more difficult one.

The chapter examines the dynamics of the library internals, and how functions compete for dominance of the library. We demonstrate that the libraries tend to converge on a small number of functions, and identify methods to test how well a library is likely to be able to scale.

It has been previously established that genetic programming can be used as an automated invention machine to synthesize designs for complex structures. In particular, genetic programming has automatically synthesized structures that infringe, improve upon, or duplicate the functionality of 21 previously patented inventions (including six 21-century patented analog electrical circuits) and has also generated two patentable new inventions (controllers). There are seven promising factors suggesting that these previous results can be extended to deliver industrial-strength automated design of analog circuits, but two countervailing factors. This chapter explores the question of whether the seven promising factors can overcome the two countervailing factors by reviewing progress on an ongoing project in which we are employing genetic programming to synthesize an amplifier circuit. The work involves a multiobjective fitness measure consisting of 16 different elements measured by five different test fixtures. The chapter describes five ways of using general domain knowledge applicable to all analog circuits, two ways for employing problem-specific knowledge, four ways of improving on previously published genetic programming techniques, and four ways of grappling with the multi-objective fitness measures associated with real-world design problems.

Traditional robust design constitutes only one step in the detailed design stage, where parameters of a design solution are tuned to improve the robustness of the system. This chapter proposes that robust design should start from the conceptual design stage and genetic programming-based open-ended topology search can be used for automated synthesis of robust systems. Combined with a bond graph-based dynamic system synthesis methodology, an improved sustainable genetic programming technique - quick hierarchical fair competition (QHFC)- is used to evolve robust high-pass analog filters. It is shown that topological innovation by genetic programming can be used to improve the robustness of evolved design solutions with respect to both parameter perturbations and topology faults.

Basic genetic programming (GP) techniques allow individuals to take advantage of some basic top-down design principles. In order to evaluate the effectiveness of these techniques, we define a design as an evolutionary frozen root node. We show that GP design converges quickly based primarily on the best individual in the initial random population. This leads to speculation of several mechanisms that could be used to allow basic GP techniques to better incorporate top-down design principles.