Hi Lee, Although converting a CPU system to GPU is not a straightforward task, I think that the theme I already intend to discuss would be highly beneficial (although not directly related) regarding the GP thing that you are asking for - Markov Chain Monte Carlo simulation. And, I sped it up tens of thousands of times (depending on the actual problem, it could be more, but also less!) comparing to the state of the art software (JAGS, Stan). Now, since the talk is not long enough to offer time for deep detail digging, I'll also look for examples that are much easier for the audience to digest - in that regard vectors and matrices are much more appropriate. Also, I intend to take special care of the stuff that if often overlooked: how to generally structure the code and do data manipulation in an efficient way on the three fronts that you need to cover: JVM, native CPU, and GPU. I'll be glad to give you pointers if you'd like to try implementing a GP hello world in ClojureCL and Neanderthal.
On Wednesday, July 13, 2016 at 11:41:28 PM UTC+2, Lee wrote: > > Dragan, > > > I would personally be interested in anything you might show about using > GPUs to speed up up genetic programming in Clojure. > > > A fair bit has been done using GPUs for GP (some can be found by searching > for GPU here > <http://liinwww.ira.uka.de/bibliography/Ai/genetic.programming.html+%3Chttp://liinwww.ira.uka.de/bibliography/Ai/genetic.programming.html>, > > but as far as I know, none of it in Clojure. > > > It would be wonderful to see a minimal example of how to take a minimal GP > system (I'd be happy to provide code) and to exploit GPUs to do bigger runs > more quickly. > > > -Lee > > On Wednesday, July 13, 2016 at 4:33:37 PM UTC-4, Dragan Djuric wrote: >> >> I'm preparing a presentation proposal for EuroClojure 2016 about Clojure >> and GPU computing, high-performance computing, data analysis, and machine >> learning. If you are interested in that area, I am open to suggestions >> about specific stuff that you would like to be covered (regardless of >> whether you plan to attend the conference itself), so I can better tailor >> the proposal to what would potentially be most interesting to the audience. >> The tools/libraries that the (proposed) talk will be based on are >> uncomplicate.org (clojurecl, neanderthal, bayadera), but I can also >> cover other aspects of the topic. >> > -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en --- You received this message because you are subscribed to the Google Groups "Clojure" group. To unsubscribe from this group and stop receiving emails from it, send an email to clojure+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
