> > The notion of a "step" seems to imply some sort of synchronous logic, > driven by a master clock. Does the cortex actually work this way? > Aside from curiosity, I'm asking because I've been thinking about the > problem of synchronizing the processing of an actor-based version of > NuPIC. This seems like it would need to force all of the actors to > wait for the slowest actors, message paths, etc. > So, I'm wondering about letting the actors operate asynchronoously, > handling and firing off messages as the scheduler (etc) allows. Of > course, this would introduce some differences in the results, but > that might not be a problem. Is this a reasonable approach to try?
My only concern with complete asynchronous activity is the idea that for a single input vector, an SDR is computed by the consideration of connections contributing to the sparsification of columns related to that single input - but the aforementioned inhibition calculation - to me this kind of implies a "synchronizing" of state prior to the inhibition calculation? What do you guys think? On Thu, Mar 5, 2015 at 7:20 PM, Tim Boudreau <[email protected]> wrote: > On Thu, Mar 5, 2015 at 7:53 PM, Rich Morin <[email protected]> wrote: > >> On Mar 5, 2015, at 16:23, Tim Boudreau <[email protected]> wrote: >> > I'd think running asynchronously would be closer to modelling the >> > biology (not to mention making it easier to leverage multiple >> > cores), but it would also mean non-deterministic results in >> > anything where the results of computation were order-dependent. >> >> Indeed. However, the main reasons I can think of for wanting a >> deterministic result would be debugging, testing, etc. I could >> use pseudo-random or even synchronous scheduling for these cases. >> > > I'd imagine where that would be problematic is if you have implementations > in multiple languages and want to prove that they actually produce the same > results. As a long as there's a knob that lets you turn the parallelism > down to 1, that ought to do the trick. > > -Tim > > -- *We find it hard to hear what another is saying because of how loudly "who one is", speaks...*
