Academia.eduAcademia.edu

Figure 4 – uploaded by Nizam Onur Sönmez

See full PDFdownloadDownload figure
Figure 5. A multi-level co-evolution / learning model for DGF. Figure 4. A mapping for the corresponding elements of Design Games Model and Design Games Framework.

Figure 5 A multi-level co-evolution / learning model for DGF. Figure 4. A mapping for the corresponding elements of Design Games Model and Design Games Framework.

Related Figures (15)

Through a series of simplification and specification operations on_ this illustration, the Design Games Model (DGM, Figure 2) has been proposec by Sénmez and Erdem (2009), with rather practical aims. In DGM, desigr processes and individual design games are clearly distinguished. There are a number of distinct tools and agents that are assumed to be reusable (indicated by letters, t and a). The agents act as carriers for techniques tha’ would guide the generic reusable tools within specific contexts. The objects that correspond to the design proposals and intermediary outputs are represented through trajectories. These outputs may be partial or holistic solution proposals in many forms (drawings, models, etc.) and are assumec to be transformed through design games.
Through a series of simplification and specification operations on_ this illustration, the Design Games Model (DGM, Figure 2) has been proposec by Sénmez and Erdem (2009), with rather practical aims. In DGM, desigr processes and individual design games are clearly distinguished. There are a number of distinct tools and agents that are assumed to be reusable (indicated by letters, t and a). The agents act as carriers for techniques tha’ would guide the generic reusable tools within specific contexts. The objects that correspond to the design proposals and intermediary outputs are represented through trajectories. These outputs may be partial or holistic solution proposals in many forms (drawings, models, etc.) and are assumec to be transformed through design games.
operators. The following mapping and evolutionary application will not be a demonstration of the applicability of the dynamism of the DGF, because such a demonstration requires, in addition to the above mentioned evolutionary strategies, further artificial intelligence techniques to carry out contextual and dynamic evaluation that demand separate studies.
operators. The following mapping and evolutionary application will not be a demonstration of the applicability of the dynamism of the DGF, because such a demonstration requires, in addition to the above mentioned evolutionary strategies, further artificial intelligence techniques to carry out contextual and dynamic evaluation that demand separate studies.
Characters hold a set of guiding parameters that stochastically determine the behavior of a tool. Most tools allow” for partial randomization, so that the exact implementation values for two applied (determined) games that use the same tool and character will be different. Therefore, instead of the generic character, these values have to be stored within the genotype of ar individual. The genotype of a candidate (i.e., individual) is a variable length chromosome, which holds a stack of initiation operations (i.e., productior games) (Figure 10). Each gene holds the implementation information of 4 specific operation, i.e., one or more tools and corresponding implementatior values. An individual is produced (genotype-phenotype mapping) by sequentially applying all determined games within the chromosome. With this genotype, it is straightforward to delete or add new genes or to apply crossover.
Characters hold a set of guiding parameters that stochastically determine the behavior of a tool. Most tools allow” for partial randomization, so that the exact implementation values for two applied (determined) games that use the same tool and character will be different. Therefore, instead of the generic character, these values have to be stored within the genotype of ar individual. The genotype of a candidate (i.e., individual) is a variable length chromosome, which holds a stack of initiation operations (i.e., productior games) (Figure 10). Each gene holds the implementation information of 4 specific operation, i.e., one or more tools and corresponding implementatior values. An individual is produced (genotype-phenotype mapping) by sequentially applying all determined games within the chromosome. With this genotype, it is straightforward to delete or add new genes or to apply crossover.
Design games: A conceptual framework for dynamic evolutionary design
Design games: A conceptual framework for dynamic evolutionary design
In two-point crossover (Figure 14), two indexes are determined in each chromosome, and the portion of chromosomes between the indexes is exchanged. The only implemented selection operator is tournament selection. Two parents have to be chosen for the generation of each couple of new candidates. In the implemented tournament selection mechanism, for each parent, two individuals are randomly selected from the population. Then
In two-point crossover (Figure 14), two indexes are determined in each chromosome, and the portion of chromosomes between the indexes is exchanged. The only implemented selection operator is tournament selection. Two parents have to be chosen for the generation of each couple of new candidates. In the implemented tournament selection mechanism, for each parent, two individuals are randomly selected from the population. Then
Figure 13. One-point crossover.
Figure 13. One-point crossover.
Several sets of trials have been carried out for the desktop icon task. Starting with only plain color areas, other types of stamps (transparent areas, gradients, pattern stamps, and texts) are gradually involved within trials. For the creation of complex stamp patterns, a two-level hierarchical approach is used (Figure 16). On the first level, a set of patterns are manually created. Through a series of evolutionary runs that use these patterns beside other stamp tools, several resulting images are obtained. Some of these images in turn were converted into stamps and added to the pool of pattern stamps for further trials. Figure 17 presents an example evolutionary run for the icon generation task. The character definition specifies a threshold for number of separate colors. The number of colors is not calculated from absolute RGB color mode values. Instead, the HSL color mode equivalent is found for each color value. The hue value is within a circular range of 0-255 points. For practical reasons, two colors are assumed same, if they are at most 20 points away from each other. In the example, if there are more than 200 separate colors, a penalty is applied. After a list of most frequent colors is found, the saturation test is applied to each, which requires the saturation of a color within specified thresholds. This example demands medium range saturation values. The evaluator character defines three tests for color areas. Each color area is given an award or a penalty, if the ratio of its area to the whole canvas is within a specified interval. The example definition prefers large (0.5 to 0.8 of canvas) and medium sized (0.3 to 0.5 of canvas) color areas, and punishes small ones (0.1 to 0.3). Color dominance is not related to color areas, but with absolute frequency of a color, which might be dispersed throughout the image. The example character rewards high frequencies and
Several sets of trials have been carried out for the desktop icon task. Starting with only plain color areas, other types of stamps (transparent areas, gradients, pattern stamps, and texts) are gradually involved within trials. For the creation of complex stamp patterns, a two-level hierarchical approach is used (Figure 16). On the first level, a set of patterns are manually created. Through a series of evolutionary runs that use these patterns beside other stamp tools, several resulting images are obtained. Some of these images in turn were converted into stamps and added to the pool of pattern stamps for further trials. Figure 17 presents an example evolutionary run for the icon generation task. The character definition specifies a threshold for number of separate colors. The number of colors is not calculated from absolute RGB color mode values. Instead, the HSL color mode equivalent is found for each color value. The hue value is within a circular range of 0-255 points. For practical reasons, two colors are assumed same, if they are at most 20 points away from each other. In the example, if there are more than 200 separate colors, a penalty is applied. After a list of most frequent colors is found, the saturation test is applied to each, which requires the saturation of a color within specified thresholds. This example demands medium range saturation values. The evaluator character defines three tests for color areas. Each color area is given an award or a penalty, if the ratio of its area to the whole canvas is within a specified interval. The example definition prefers large (0.5 to 0.8 of canvas) and medium sized (0.3 to 0.5 of canvas) color areas, and punishes small ones (0.1 to 0.3). Color dominance is not related to color areas, but with absolute frequency of a color, which might be dispersed throughout the image. The example character rewards high frequencies and
Figure 17. An example run for the icon generation task.
Figure 17. An example run for the icon generation task.
Related topics:
Design TheoryComputational DesignDesign AutomationEvolutionary designAI in Design
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved