Hi Ralf - Thanks for the thoughtful response.
In order to make your port work with node, (actually io.js, which is a fork of node, soon to be merged back into node) - I had to make some changes. I suspect that my project will, over time, diverge substantially from your project, since the goals are a bit different. Going forward, I will keep an eye on your project and merge what I can. Thank you so much for getting the project to where it is - it would have taken me a long time to accomplish what you did! And yes - it does illuminate the usefulness of a statically typed language, such as Java. Cheers, Jeff On Mon, May 18, 2015 at 9:18 AM Ralf Seliger <[email protected]> wrote: > Hey Jeff, > > sorry for the late reply, I was out of town for a couple of days. > > Am 13.05.2015 um 22:31 schrieb Jeff Fohl: > > Hello Ralf - > > > > You might remember from the nupic-hackers mailing list, that I am also > > working on an implementation of HTM in Javascript. I have found that > > looking at your port, htm.JavaScript, is helping me understand some > > concepts. To better understand things, I forked htm.JavaScript and > started > > tinkering. My preferred JS environment is node.js, so I have been working > > on converting htm.JavaScript to work with node.js. > > > > Do you have any interest in getting this to work in node.js? Also, do you > > wish to have any help with htm.JavaScript? > > > > I am also interested to hear what your larger goals are for > htm.JavaScript. > > > > Best regards, > > > > Jeff Fohl > Actually my focus is on running NuPIC in web browsers. Eventually, I'd > like to create a distributed system/environment where various HTM/CLA > controlled "entities" interact with the environment as well as which > each other (so the sensori-motor stuff Numenta is researching right now > would have to be in place, too). > > So the larger goal is some kind of "artificial life platform" where in > the end I hope to see some sort of emerging behaviour. Of course, there > would have to be some evolutionary stuff going on, as well. But one > would *not* need any artificial performance functions like in classical > genetic algorithms. The only hard-coded goal would be to reproduce, > everything else ought to follow/emerge from that due to the learning > abilities of those entities. Once development of the interaction > software starts, node.js might come into play, but that is somewhere > beyond the horizon. > > I don't think I need help right now. At this point it's mostly a > straight forward translation job, although one learns to appreciate the > virtues of a statically typed language like Java. Anyway, thanks for the > offer, if you have any suggestions for improvements, feel free to issue > a pull request. At the very least it would give me an opportunity to > learn how that works ;-) > > RS > >
