![]() |
||
|
Seminar Abstract 02 Mar 2007: There is a growing body of research concerned with the use of machines to facilitate or exhibit "creative processes". A fundamental aspect of such endeavours is the generation of novel structures by machines. Interactive Evolutionary Computation is a seemingly promising approach wherein the user typically breeds pleasing structures to generate new - hopefully better - ones over a number of generations/cycles. This talk will present initial results from using a form of Evolution Strategy for such work. In particular, the algorithm is extended to search spaces of varying complexity (length), exploiting the concept of parameter self-adaptation inherent to Evolution Strategies. This is potentially powerful as self-adaptation enables the way in which the solution space is explored to be tailored to the user as they search. Application to a very simple abstract art tool will be shown. Time permitting the use and training of non-linear media,
such as in vitro neuronal networks, as a new approach to the generation
of interesting structures will also be touched upon very briefly. |
||