Name
Abstract | ||
|---|---|---|
Interactive Evolutionary Computation (IEC) faces a dilemma between search performance and cost due to user fatigue during solution evaluation. This becomes more serious in the case of real-world problems as they generally require the optimization of both qualitative and quantitative objective functions. Although user preference prediction is a promising way, it requires a certain training data obtained from users. On the other hand, little attention has been given to a search role assignment between a user and system, which essentially should be adjusted according to user's domain knowledge, search progress, etc. Therefore, this paper proposes a method called cooperative evolution by user and system (CEUS), which allows a user to change roles dynamically, and it learns user preferences even from a few user operations, thereby alleviating user fatigue even for problems involving both qualitative and quantitative objective functions. The proposed CEUS is applied to a Japanese anagram generation problem, and experiments in the problem showed that the proposed CEUS allowed a user to dynamically define roles and change operation timing and that it well supports users' convergent and divergent thinking. |
Year | DOI | Venue |
|---|---|---|
2012 | 10.1145/2330784.2331030 | GECCO (Companion) |
Keywords | Field | DocType |
search role assignment,user preference,proposed ceus,japanese anagram generation problem,search performance,user fatigue,search progress,quantitative objective function,user preference prediction,user operation,user-system cooperative evolution,japanese anagram sentence generation,objective function,interactive evolutionary computation,domain knowledge | Interactive evolutionary computation,Mathematical optimization,User experience design,Convergent thinking,Domain knowledge,Computer science,Anagram,User modeling,Artificial intelligence,Computer user satisfaction,User requirements document,Machine learning | Conference |
Citations | PageRank | References |
3 | 0.45 | 1 |
Authors | ||
2 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Satoshi Ono | 1 | 219 | 39.83 |
| Shigeru Nakayama | 2 | 75 | 16.14 |