Is it advisable to use Genetic Algorithm for a system that is required to give a real time response where the response is to be computed using GA? Please, share your ideas.
In general, I would say no. GAs usually take a while to converge to a really good solution.
BUT, you can always terminate the GA and give the best solution found so far at any time and after any number of generations (be it large or small). In that case, you would have to tune the parameters (specially the popluation size which should not be very large if you want it to converge fast -although this really depends on the difficulty of the problem-, but also crossover, etc) so that the algorithm has already found something decent (possibly a local optimum that is okay for you) after only a few generations. Or even adapt the values of the parameters when you know the time limit is getting near, in order to enhance exploitation (and forget exploration) for a while before terminating. Maybe evolution strategies (ES) other than GAs work better for your case - try Googling "evolutionary computation for real time" or something like that.
It depends on your definition of real-time. Are you talking about an instantaneous answer to a users request, or are you referring to some (automatized) process which requires an answer within a specific time scale? If you want to the keep the answer of the GA within a time window, this window just defines your target precision of the GA.
@Marcus Neuer, Thank you for contributing. I meant a spontaneous response to a user;s request.
@Shampa Chakraverty, Thank you for your ideas. Am building a desktop application and don't have GPU infrastructure. The system is a firm RT system.
Genetic algorithms (GAs) are well-suited for real time system.
- Badges
- Science topic
- Similar topics
- Agricultural Science
- Agronomy