PhD Proposal

Creative Design Using Collaborative Interactive Genetic Algorithms :Creative Design Using Collaborative Interactive Genetic Algorithms Juan C. Quiroz PhD Dissertation Proposal Friday, September 5, 2008 Evolutionary Computing Systems Lab Department of Computer Science & Engineering


Acknowledgements :2 Acknowledgements Dr. Sushil Louis Dr. Sergiu Dascalu Dr. Swatee Naik Dr. Bobby Bryant Dr. Amit Banerjee Dr. Darren Platt


Outline :3 Outline Creativity in design Computational model of creativity Interactive evolutionary computation Collaboration in interactive evolutionary computation Proposal of future work


Creativity :4 Creativity Creativity is a mental process involving the generation of new ideas or concepts, or new associations of the creative mind between existing ideas or concepts. (Wikipedia) Novel and useful 134139834


Motivation :5 Motivation Design process Conceptual design Detailed design Evaluation Iterative redesign


Conceptual Design :6 Conceptual Design Subjective evaluation of alternative design concepts Aesthetics and other subjective criteria What is the formula for how designers evaluate subjective criteria? Collaborative design


Our Goal :7 Our Goal Create a computational model of creative design Allows for subjective evaluation Supports collaboration Has the potential to produce creative content Hypothesis Collaborative interactive genetic algorithms are a viable computational model of creative design


Previous Work :8 Previous Work Used interactive genetic algorithms for user interface design (2007) Explored techniques to reduce user fatigue in interactive genetic algorithms (2007) Developed and proposed a computational model of creative design (2008)


Genetic Algorithms :9 Genetic Algorithms Population based search technique Natural selection Survival of the fittest


Interactive Genetic Algorithms (IGAs) :10 Interactive Genetic Algorithms (IGAs) Fashion design (Kim 2000) Micromachine design (Kamalian 2005) Music, editorial design (Takagi 2001) Traveling salesman problem (Louis 1999)


User Fatigue in IGAs :11 User Fatigue in IGAs GAs tend to rely on: Large populations Many generations Suboptimal solutions Noisy fitness landscapes


Alleviating User Fatigue :12 Alleviating User Fatigue Use small population sizes Display a subset of population (Lee 1999) Accelerate convergence through prediction (Llora 2005)


IGAS in UI Design :13 IGAS in UI Design Goals Use a subset of population for user evaluation (Takagi 2001) Ask user to pick the best and worst instead of evaluating all individuals Incorporate objective heuristics into IGA (Kamalian et al. 2005) Evaluate and assess how user evaluation every tth would affect IGA performance (Kamalian et al. 2005)


Experimental Setup :14 Experimental Setup Greedy simulated user 30 independent runs, 200 generations Population size of 100 Display method comparison: best 10, random 10, best 5 and worst 5 User input every 1, 5, 10, 20, 40, 80 generations


Who from the population do we display for user evaluation? :15 Who from the population do we display for user evaluation? Fitness Convergence Convergence to Blue UIs


How often do we ask for user input? :16 How often do we ask for user input? User input every tth generation Fitness convergence High values of t Low values of t


How often do we ask for user input? :17 How often do we ask for user input? User input every tth generation Convergence to blue UIs High values of t Low values of t


Experimental Setup: Actual Users :18 Experimental Setup: Actual Users Three users 30 generations Pick the one they like the best and the one they like the least 5 sessions User input every 1, 3, 5, 10, 15 generations 30, 10, 6, 3, and 2 user evaluations respectively


Results :19 Results


What leads to the drop in average performance? :20 What leads to the drop in average performance? Two sessions with a user Comparison to user selected worst turned on Comparison to user selected worst turned off Always pick the same UI as the best Ask for user input every 3 generations


What leads to the drop in average performance? :21 What leads to the drop in average performance?


IGAs for UI Design Conclusions :22 IGAs for UI Design Conclusions We can use IGAs to evolve UIs Our simulated user and actual users are able to effectively bias the evolution of UIs UIs reflect coded guidelines of style Reduce user fatigue Interpolation technique Asking for less user input


Computational Model of Creative Design :23 Computational Model of Creative Design Collaborative Interactive Genetic Algorithm


Design Space Exploration :24 Design Space Exploration


IGAP: Interactive GA Peer to Peer :25 IGAP: Interactive GA Peer to Peer


Individual Visualization :26 Individual Visualization Bedroom Eating area Bathroom


Collaborative Visualization :27 Collaborative Visualization


Experimental Setup :28 Experimental Setup Design a floorplan for a two-bedroom, one-bathroom apartment Living room should face north-west The two bedrooms should not have a common wall At least one of the bedrooms should have direct access to the bathroom


Experimental Setup :29 Experimental Setup Authors and two colleagues evolved floorplans Individually Collaboratively Ten computer science graduate students evaluated the designs by taking a survey The plans were evaluated for creative content based on practicality and originality


Floorplan Results :30 Floorplan Results


Results :31 Results


Computational Model of Creativity Conclusions :32 Computational Model of Creativity Conclusions IGAP: Interactive Genetic Algorithm Peer to Peer Collaborative interactive genetic algorithm Floorplans developed via collaboration were more original than those created individually


Proposal :33 Proposal Validation of creativity model User modeling with machine learning Explicit expansion of the search space via masking


Validation of Creativity Model :34 Validation of Creativity Model User studies Design participants Design creation individually and collaboratively Evaluation participants Score designs created by design participants Case studies Floorplanning Blogging page Brochure


User Modeling :35 User Modeling Can we learn the design style of a user by observing his/her selections? Is this sufficient data? SVM, Neuroevolution, Weka


IGAP :36 IGAP User Modeling User Modeling User Modeling


Creativity via Masking :37 Creativity via Masking


Creativity via Masking :38 Creativity via Masking After Peer Injection


Contributions :39 Contributions A novel computational model of creativity based on collaborative IGAs Model supports and reflects the collaborative methodology followed in design teams A model that can improve the quality of designs


Publications :40 Publications J.C. Quiroz, S.M. Dascalu, and S.J. Louis, “Human guided evolution of XUL user interfaces,” CHI '07 extended abstracts on Human factors in computing systems, San Jose, CA, USA: ACM Press, 2007, pp. 2621-2626. J.C. Quiroz, S.J. Louis, and S.M. Dascalu, “Interactive evolution of XUL user interfaces,” Proceedings of the 9th annual conference on Genetic and evolutionary computation, London, England: ACM Press, 2007, pp. 2151-2158. J. Quiroz et al., “Interactive Genetic Algorithms for User Interface Design,” Evolutionary Computation, 2007. CEC 2007. IEEE Congress on, 2007, pp. 1366-1373. J. Quiroz et al., “Software Environment for Research on Evolving User Interface Designs,” Software Engineering Advances, 2007. ICSEA 2007. International Conference on, 2007, p. 84. A. Banerjee, J.C. Quiroz, and S.J. Louis, “A Model of Creative Design Using Collaborative Interactive Genetic Algorithms,” Design Computing and Cognition '08, Springer, 2008, pp. 397-416. J.C. Quiroz, A. Banerjee, and S.J. Louis, “IGAP: interactive genetic algorithm peer to peer,” Proceedings of the 10th annual conference on Genetic and evolutionary computation, Atlanta, GA, USA: ACM, 2008, pp. 1719-1720.


Comments/Questions? :41 Comments/Questions? quiroz@cse.unr.edu www.cse.unr.edu/~quiroz Acknowledgements This work was supported in part by contract number N00014-0301-0104 from the Office of Naval Research and the National Science Foundation under Grant no. 0447416.


Slide 42 :42


Representation is Key :43 Representation is Key


IGA Operators :44 IGA Operators


UI Representation :45 UI Representation Two chromosomes Widget chromosome Layout chromosome


Genetic Operators for UI IGA :46 Genetic Operators for UI IGA Single point crossover Bit flip mutation PMX – partial mapped crossover Swap mutation


Widget Color :47 Widget Color RGB color model Red = (255, 0, 0), Green = (0, 255, 0), Blue = (0, 0, 255) 224 color space for each widget HSV Same gamut as RGB No significant efficiency difference in RGB and HSV


Fitness Evaluation :48 Fitness Evaluation Ask the user to select the best and worst UIs from the subset displayed Interpolate the subjective fitness of individuals in population Compute the objective metrics taken from guidelines of style Fitness = w1 * subjective + w2 * objective


Fitness Evaluation: Step 1 :49 Fitness Evaluation: Step 1 Best Worst


Subjective Fitness Interpolation :50 Subjective Fitness Interpolation Compare to best Compare to worst


Objective Fitness Computation :51 Objective Fitness Computation High contrast between widget colors and background color Low contrast between widget colors


Slide 52 :52 Fitness = w1 * subjective fitness + w2 * objective fitness