Andrés Gómez de Silva Garza

A Cognitive Evaluation of a Computer System for Generating Mondrian-Like Artwork

In this paper we present a computer system based on the notion of evolutionary algorithms that, without human intervention, generates artwork in the style of the Dutch painter Piet Mondrian. Several implementation-related decisions that have to be made in order to program such a system are then discussed. The most important issue that has to be considered when implementing this type of system is the subroutine for evaluating the multiple potential artworks generated by the evolutionary algorithm, and our method is discussed in detail. We then present the set-up and results of a cognitive experiment that we performed in order to validate the way we implemented our evaluation subroutine, showing that it makes sense. Finally, we discuss our results in relation to other research into the computer generation of artwork that fits particular styles existing in the real world.

Case-Based art

While there have been plenty of applications of case-based reasoning (CBR) to different design tasks, rarely has the methodology been used for generating new works of art. If the goal is to produce completely novel artistic styles, then perhaps other reasoning methods offer better opportunities for producing interesting artwork. However, if the goal is to produce new artwork that fits a previously-existing style, then it seems to us that CBR is the ideal strategy to use. In this paper we present some ideas for integrating CBR with other artificial intelligence techniques in order to generate new artwork that imitates a particular artistic style. As an example we show how we have successfully implemented our ideas in a system that produces new works of art in the style of the Dutch painter Piet Mondrian. Along the way we discuss the implications that a task of this nature has for CBR and we describe and provide the results of some experiments we performed with the system.

An Evolutionary Process Model for Design Style Imitation

In this paper we propose a process model for producing novel constructs that fall within an existing design or artistic style. The process model is based on evolutionary algorithms. We then present an implementation of the process model that we used in order to imitate the style of the Dutch painter Mondrian. Finally we explain and give the results of a cognitive experiment designed to determine the effectiveness of the process model, and provide a discussion of these results.

Automating Evolutionary Art in the Style of Mondrian

Many recent systems for computer artwork generation with evolutionary algorithms have been interactive, relegating the task of evaluating new genotypes to their users. In contrast, we are interested in fully automating the entire evolutionary cycle applied to art work generation. We set out to produce a program that would generate artwork in the style of the Dutch painter Piet Mondrian, who was active mainly in the first half of the 20th century. He developed his own distinctive and abstract style (called simply de stijl, which is Dutch for the “style”). Paintings in Mondrian’s style typically include vertical and horizontal black lines over a white background, with some or all of the primary colors (blue, red, and yellow), plus black, filling in some of the square or rectangular regions (or parts of the regions) separated out from the background by the black lines. It is this style that our system tries to emulate.

An introduction to and comparison of computational creativity and design computing

The interrelated fields of computational creativity and design computing, sometimes also referred to as design science, have been gaining momentum over the past two or three decades. Many frequent international conference series, as well as more sporadic stand-alone academic events, have emerged to prove this. As maturing fields, it is time to take stock of what has come before and try to come up with a cohesive description of the theoretical foundations and practical advances that have been made. This paper presents such a description in the hope that it helps to communicate what the fields are about to people that are not directly involved in them, hopefully drawing some of them in.The interrelated fields of computational creativity and design computing, sometimes also referred to as design science, have been gaining momentum over the past two or three decades. Many frequent international conference series, as well as more sporadic stand-alone academic events, have emerged to prove this. As maturing fields, it is time to take stock of what has come before and try to come up with a cohesive description of the theoretical foundations and practical advances that have been made. This paper presents such a description in the hope that it helps to communicate what the fields are about to people that are not directly involved in them, hopefully drawing some of them in.

Exploring the sensitivity to representation of an evolutionary algorithm for the design of shapes

Evolutionary algorithms have been used in many ways for design tasks and other creative endeavors. The choice of representation in these evolutionary algorithms influences the way in which the operators used by a given evolutionary algorithm need to be implemented. In this paper, I explore the way in which it also influences the performance of the algorithm, both from the point of view of the reliability and speed of the algorithms as well as from the point of view of the designs generated by them.

A coevolutionary method for automating airport gate scheduling

The problem of assigning gates to aircraft that are due to arrive at an airport is one that involves a dynamic task environment. Airport gates can only be assigned if they are currently available, but deciding which gate to assign to which flight also involves satisfying multiple additional constraints. Once a solution has been found, new incoming flights will have approached the airspace of the airport in question, and these will require arrival gates to be assigned to them, so the entire process must be repeated. These observations have led us to propose a coevolutionary model for automating the airport gate scheduling problem. We represent the genotypes of two species. One species corresponds to the current problem to be addressed (a list of departing and arriving flights at a given time-step) and the other species corresponds to the solutions being proposed for that problem (a list of possible gate assignments for the arriving flights). An evolutionary algorithm which operates on a population of solution genotypes is used to solve each instance of the airport gate scheduling problem. A coevolutionary algorithm in which the two species influence each other, which incorporates the previously-mentioned evolutionary algorithm once at each time-step, models the fact that multiple instances of the problem occur over time as an airport operates.

Modifying the fitness function during the use of an evolutionary algorithm for design

We use an evolutionary algorithm in which we change the fitness function periodically to model the fact that objectives can change during creative problem solving. We performed an experiment to observe the behavior of the evolutionary algorithm regarding its response to these changes and its ability to successfully generate solutions for its creative task despite the changes. An analysis of the results of this experiment sheds some light into the conditions under which the evolutionary algorithm can respond with varying degrees of robustness to the changes.

Comparing Different Implementations of the Same Fitness Criteria in an Evolutionary Algorithm for the Design of Shapes

Evolutionary algorithms (EAs) have been used in varying ways for design and other creative tasks. One of the main elements of these algorithms is the fitness function used by the algorithm to evaluate the quality of the potential solutions it proposes. The fitness function ultimately represents domain knowledge that serves to bias, constrain, and guide the algorithm’s search for an acceptable solution. In this paper, we explore the degree to which the fitness function’s implementation affects the search process in an evolutionary algorithm. To perform this, the reliability and speed of the algorithm, as well as the quality of the designs produced by it, are measured for different fitness function implementations. These measurements are then compared and contrasted.

Commonsense Knowledge as the Glue in a Hybrid Model of Computational Creativity

In this paper, we describe on-going work on the ensemble of two models for computational creativity using commonsense knowledge. The GENCAD model of computational creativity proposes the use of an evolutionary algorithm (EA) that employs a population of exemplars as a starting point for its search, unlike traditional EAs, which employ a randomly-generated initial population. The EA, operating on this population, is then used to generate new potentially creative solutions. GENCAD has been instantiated in the domains of structural design of tall buildings and feng shui-compliant residential floor-plan design. The MEXICA model of computational creativity also begins with a set of exemplars as a starting point, but it analyzes these exemplars based on a domain theory. The generalized model that is obtained from analyzing the set of exemplars is then used to guide the generation of new solutions. MEXICA has been instantiated in the domain of plot generation for stories involving themes, characters and locations from the Mexica culture of ancient Mexico. In the hybrid model, we combine the two original models to generate plots for stories of the same sort that MEXICA generates, but using GENCADs process model to do so. To confirm that the generated plots make sense, the hybrid model integrates commonsense knowledge into its evaluation module.