Hello! I don't know if you mean by toolkit only programs in python but I am writing a program to create and use a HTM-network in Matlab. It basically works like in this paper [1], but with some functions added to speed up performance (which is still very slow). My encoder takes scalar input for the moment, but I want to improve it as soon as the rest works well. I also have some questions but I will ask them under another topic. If somebody wants to try/use my code, just send me an email, but first I have to put it a little in order and write some more documentation.
Thanks for sharing your great ideas with everybody! Andreas [1] http://arxiv.org/ftp/arxiv/papers/1511/1511.00083.pdf Am 13.01.2016 um 18:28 schrieb Sebastián Narváez <[email protected]>: > Not sure if you can call this a framework, but I did a Layer class to > experiment with different Encoders, SPs and TMs combinations. The class > LearningModels uses the Layer class, so you can get a grasp on how I've been > using it. Note that this was just a rapid solution and has not been heavily > tested. > > https://github.com/larvasapiens/htm-teul/tree/master/Learning > > On Wed, Jan 13, 2016 at 11:52 AM, cogmission (David Ray) > <[email protected]> wrote: > Here I was, swimming peacefully along in my pond when out of nowhere, this > baited hook appeared in front of my face. That worm looks awful tasty too! :-) > > Cheers, > David > > On Wed, Jan 13, 2016 at 1:18 AM, Raf <[email protected]> wrote: > Hello NuPIC, > > I'm personally trying to do my best to replicate most of what I know about > the neocortex from my neurophysiology and neurology classes into my own > elaboration of the NuPIC core. To start with I was very ambitious: EPSP, > IPSP, proximal and distal dendrites classes, chemical receptor classes, > auto-expanding cell/columns. > > Unfortunately as a programmer I'm pretty slow and I'm sure a better coder > would make a better job. But as soon as I'm done -if it is worth it- I'll > definitely share it with you guys under a AGPL license. > > Thanks for NuPIC, I really love the work you've done. > > Raf > > > > On 13/01/2016 07:57, Matthew Taylor wrote: >> Hello NuPIC, >> >> I have heard about some of you guys and gals writing your own NuPIC >> frameworks either by using SP and TM and classifiers together manually, or >> using the Network API. That's great! How about sharing them with the rest of >> the community? I would love to have some new code to experiment with besides >> the OPF. (I've even pondered going over the the Dark Side (JVM OMG)) >> >> <Mail-Anhang.jpeg> >> >> Please share your work under an AGPL license. >> >> --------- >> Matt Taylor >> OS Community Flag-Bearer >> Numenta > > -- > Raf > > www.madraf.com/algotrading > reply to: [email protected] > skype: algotrading_madraf > > > > -- > With kind regards, > > David Ray > Java Solutions Architect > > Cortical.io > Sponsor of: HTM.java > > [email protected] > http://cortical.io >
