I can do a short talk about "things to do to put NUPIC in production . Any thoughts?
On Wed, Oct 28, 2015 at 10:17 AM, Matthew Taylor <[email protected]> wrote: > There will be whiteboards.... :) > --------- > Matt Taylor > OS Community Flag-Bearer > Numenta > > > On Wed, Oct 28, 2015 at 7:58 AM, Richard Crowder <[email protected]> wrote: > > Agree. The recent informal meetup was great. And looking likely to be a > > format used more. As well as more specific meets, such as the NuPIC > > Development Progress Reviews and Office Hours. > > > > On Wed, Oct 28, 2015 at 2:33 PM, Felix Andrews <[email protected]> wrote: > >> > >> No I was just talking about the idea of a special forum session, to beat > >> out some precise computational experiments from the theory. But I now > think > >> we shouldn't do that. Just leave it ad-hoc. > >> > >> It's often the incidental conversations that are most valuable, so I was > >> suggesting we don't fill up all the time with formal talks. I realise > this > >> leaves out remote people but there it is. Maybe we could hire some > >> telepresence robots. > >> > >> > >> On 28 October 2015 at 20:20, Richard Crowder <[email protected]> wrote: > >>> > >>> I gained the impression that it was split into two schedules for the > >>> community side. Having longer presentations with questions and answers > after > >>> or throughout. Then shorter time-limited 'lightning' talks with time > limited > >>> Q&A after (potentially before the longer talks). To help presenters > enter > >>> videos and schedule a time for them to attend and answer questions, > time > >>> zone differences etc. > >>> I'm guessing your not referring to the lightning talk side with your > >>> second thought? > >>> > >>> On Wed, Oct 28, 2015 at 11:51 AM, Felix Andrews <[email protected]> > wrote: > >>>> > >>>> On second thought, I don't think we need a scheduled, structured > >>>> session. It would be better to have ad-hoc discussions. Just don't > fill up > >>>> all the time with talks. > >>>> > >>>> > >>>> On Wednesday, 28 October 2015, Felix Andrews <[email protected]> wrote: > >>>>> > >>>>> > >>>>> On 27 October 2015 at 12:31, Matthew Taylor <[email protected]> > wrote: > >>>>>> > >>>>>> I nominate David Ray to moderate. (If he wants to, of course). > >>>>>> > >>>>> Seconded! > >>>>> > >>>>> Any ideas on how to structure this kind of session? I'm sure there > will > >>>>> be plenty of ideas raised by Sergey, Fergal and others. It might be > obvious > >>>>> what to focus on. Or not. > >>>>> > >>>>> I find that putting ideas into code exposes a lot of unspecified > >>>>> details, which may be very important in their effects. So personally > I would > >>>>> like us to propose some concrete mechanisms, avoiding hand waving as > much as > >>>>> possible (sure, we need hand waving sometimes). It would be good to > get > >>>>> those people with more neuroscience knowledge involved at that > detailed > >>>>> level, I think. > >>>>> > >>>>> > >>>>> > >>>>>> > >>>>>> On Oct 26, 2015, at 8:20 PM, Felix Andrews <[email protected]> wrote: > >>>>>> > >>>>>> Sergey, I would love to hear that talk. Distilled neuroscience is > >>>>>> gold. > >>>>>> > >>>>>> Perhaps, along with Fergal's talk, it could lead into a > >>>>>> forum/discussion on potential computational experiments inspired by > >>>>>> neuroscience? An explicit bridge-building exercise between theory > and > >>>>>> practice. We would need a skilled moderator. > >>>>>> > >>>>>> > >>>>>> On 27 October 2015 at 07:43, Marcus Lewis <[email protected]> > wrote: > >>>>>>> > >>>>>>> +1, I'd be interested in your talk. > >>>>>>> > >>>>>>> On Mon, Oct 26, 2015 at 4:40 PM Sergey Alexashenko > >>>>>>> <[email protected]> wrote: > >>>>>>>> > >>>>>>>> Hey David, > >>>>>>>> > >>>>>>>> Sorry, I must have not been clear. I do not intend to present > >>>>>>>> theories from neuroscience, just evidence. For example, we now > know that > >>>>>>>> inhibitory neurons can block off information input to an > excitatory neuron's > >>>>>>>> cell body from one specific dendrite. That seems like it could be > important, > >>>>>>>> but HTM has no way to account for that. There are a few other > similar > >>>>>>>> mechanics. > >>>>>>>> > >>>>>>>> Obviously, I have theories about these behaviors (and so do other > >>>>>>>> people), but I do not want to present them for similar reasons to > those in > >>>>>>>> your email. > >>>>>>>> > >>>>>>>> What are your thoughts on that approach? > >>>>>>>> > >>>>>>>> Sergey > >>>>>>>> > >>>>>>>> On Mon, Oct 26, 2015 at 4:18 PM, David Ray > >>>>>>>> <[email protected]> wrote: > >>>>>>>>> > >>>>>>>>> Sergey, > >>>>>>>>> > >>>>>>>>> Can you put it in context - show how the contrary opinions > actually > >>>>>>>>> demonstrate computational viability - and not just someone's > opinions? Show > >>>>>>>>> how the challenging theories have a computational model which > illustrates a > >>>>>>>>> more efficient likelihood of prediction accuracy? Because If all > the > >>>>>>>>> challenging theories do is propose a theory, then how is it a > contribution > >>>>>>>>> considering HTM theory yields actual results. > >>>>>>>>> > >>>>>>>>> So personally I'm not interested unless the theory can > demonstrate > >>>>>>>>> an increase in cognitive resolution. > >>>>>>>>> > >>>>>>>>> But that's just me... > >>>>>>>>> > >>>>>>>>> Sent from my iPhone > >>>>>>>>> > >>>>>>>>> On Oct 26, 2015, at 5:28 PM, Sergey Alexashenko > >>>>>>>>> <[email protected]> wrote: > >>>>>>>>> > >>>>>>>>> Hey everyone, > >>>>>>>>> > >>>>>>>>> Do you think there would be any interest in a short talk on some > >>>>>>>>> neuroscience challenges to HTM theory? I am specifically > thinking about > >>>>>>>>> dendritic computation, role of inhibitory neurons and non-Hebbian > >>>>>>>>> plasticity. I know this sounds dense, but I am reasonably sure I > can make it > >>>>>>>>> sound accessible to people who have only On Intelligence and the > white paper > >>>>>>>>> as their background. > >>>>>>>>> > >>>>>>>>> Sergey > >>>>>>>>> > >>>>>>>>> On Oct 26, 2015 3:22 PM, "Fergal Byrne" > >>>>>>>>> <[email protected]> wrote: > >>>>>>>>>> > >>>>>>>>>> Yes, either Matt will get a recording, or the speaker will, or > >>>>>>>>>> someone will come and use their own camera/mike. Or best, all > of these. > >>>>>>>>>> > >>>>>>>>>> On Mon, Oct 26, 2015 at 9:44 PM, Marcus Lewis < > [email protected]> > >>>>>>>>>> wrote: > >>>>>>>>>>> > >>>>>>>>>>> Will these talks be recorded? I vote for "yes". > >>>>>>>>>>> > >>>>>>>>>>> On Mon, Oct 26, 2015 at 1:23 PM, Matthew Taylor > >>>>>>>>>>> <[email protected]> wrote: > >>>>>>>>>>>> > >>>>>>>>>>>> Thanks Fergal, > >>>>>>>>>>>> > >>>>>>>>>>>> Please edit this wiki document: > >>>>>>>>>>>> > >>>>>>>>>>>> > https://github.com/numenta/nupic/wiki/HTM-Challenge-Community-Meetup-Planning > >>>>>>>>>>>> > >>>>>>>>>>>> --------- > >>>>>>>>>>>> Matt Taylor > >>>>>>>>>>>> OS Community Flag-Bearer > >>>>>>>>>>>> Numenta > >>>>>>>>>>>> > >>>>>>>>>>>> > >>>>>>>>>>>> On Mon, Oct 26, 2015 at 12:54 PM, Fergal Byrne > >>>>>>>>>>>> <[email protected]> wrote: > >>>>>>>>>>>> > Thanks Matt for organising the Hangout this > morning/afternoon. > >>>>>>>>>>>> > > >>>>>>>>>>>> > The Community Meetup on Friday, November 13th [1] is hosted > by > >>>>>>>>>>>> > Numenta but > >>>>>>>>>>>> > is our day. Please have your say in designing and shaping > how > >>>>>>>>>>>> > the day works. > >>>>>>>>>>>> > > >>>>>>>>>>>> > The current proposal is a mixture of longish talks > (40-50mins > >>>>>>>>>>>> > plus Q&A) and > >>>>>>>>>>>> > lightning talks (5-15mins plus Q&A). > >>>>>>>>>>>> > > >>>>>>>>>>>> > We have so far pencilled in two long-form talks: > >>>>>>>>>>>> > > >>>>>>>>>>>> > 1. Donna Dubinsky (Numenta co-founder and CEO) on the > >>>>>>>>>>>> > "Business of HTM" > >>>>>>>>>>>> > 2. Me on "Symphony from Synapses: Neocortex as a Universal > >>>>>>>>>>>> > Dynamical Systems Modeller" > >>>>>>>>>>>> > > >>>>>>>>>>>> > We have from 3pm until 9ish, so there's plenty of room for > >>>>>>>>>>>> > talks of both > >>>>>>>>>>>> > types. I'd personally enjoy a good mixture of a few more > >>>>>>>>>>>> > long-form and a > >>>>>>>>>>>> > good number of swift talks. > >>>>>>>>>>>> > > >>>>>>>>>>>> > Matt is going to set up a Wiki page for this (he'll post a > >>>>>>>>>>>> > link here), so > >>>>>>>>>>>> > please jump in with your talk proposals - specify approx > >>>>>>>>>>>> > length of the talk > >>>>>>>>>>>> > so we can figure out how to schedule them. > >>>>>>>>>>>> > > >>>>>>>>>>>> > For those who cannot attend in person, we'll hopefully > >>>>>>>>>>>> > organise a > >>>>>>>>>>>> > webcast/hangout way to deliver your contribution. > >>>>>>>>>>>> > > >>>>>>>>>>>> > This is all subject to objections/amendments from you guys, > >>>>>>>>>>>> > nothing is set > >>>>>>>>>>>> > in stone. > >>>>>>>>>>>> > > >>>>>>>>>>>> > Regards, > >>>>>>>>>>>> > > >>>>>>>>>>>> > Fergal Byrne > >>>>>>>>>>>> > > >>>>>>>>>>>> > [1] http://www.meetup.com/numenta/events/224711563/ > >>>>>>>>>>>> > > >>>>>>>>>>>> > -- > >>>>>>>>>>>> > > >>>>>>>>>>>> > Fergal Byrne, Brenter IT @fergbyrne > >>>>>>>>>>>> > > >>>>>>>>>>>> > http://inbits.com - Better Living through Thoughtful > >>>>>>>>>>>> > Technology > >>>>>>>>>>>> > http://ie.linkedin.com/in/fergbyrne/ - > >>>>>>>>>>>> > https://github.com/fergalbyrne > >>>>>>>>>>>> > > >>>>>>>>>>>> > Founder of Clortex: HTM in Clojure - > >>>>>>>>>>>> > https://github.com/nupic-community/clortex > >>>>>>>>>>>> > Co-creator @OccupyStartups Time-Bombed Open License > >>>>>>>>>>>> > http://occupystartups.me > >>>>>>>>>>>> > > >>>>>>>>>>>> > Author, Real Machine Intelligence with Clortex and NuPIC > >>>>>>>>>>>> > Read for free or buy the book at > >>>>>>>>>>>> > https://leanpub.com/realsmartmachines > >>>>>>>>>>>> > > >>>>>>>>>>>> > e:[email protected] t:+353 83 4214179 > >>>>>>>>>>>> > Join the quest for Machine Intelligence at > http://numenta.org > >>>>>>>>>>>> > Formerly of Adnet [email protected] http://www.adnet.ie > >>>>>>>>>>>> > >>>>>>>>>>> > >>>>>>>>>> > >>>>>>>>>> > >>>>>>>>>> > >>>>>>>>>> -- > >>>>>>>>>> > >>>>>>>>>> Fergal Byrne, Brenter IT @fergbyrne > >>>>>>>>>> > >>>>>>>>>> http://inbits.com - Better Living through Thoughtful Technology > >>>>>>>>>> http://ie.linkedin.com/in/fergbyrne/ - > >>>>>>>>>> https://github.com/fergalbyrne > >>>>>>>>>> > >>>>>>>>>> Founder of Clortex: HTM in Clojure - > >>>>>>>>>> https://github.com/nupic-community/clortex > >>>>>>>>>> Co-creator @OccupyStartups Time-Bombed Open License > >>>>>>>>>> http://occupystartups.me > >>>>>>>>>> > >>>>>>>>>> Author, Real Machine Intelligence with Clortex and NuPIC > >>>>>>>>>> Read for free or buy the book at > >>>>>>>>>> https://leanpub.com/realsmartmachines > >>>>>>>>>> > >>>>>>>>>> e:[email protected] t:+353 83 4214179 > >>>>>>>>>> Join the quest for Machine Intelligence at http://numenta.org > >>>>>>>>>> Formerly of Adnet [email protected] http://www.adnet.ie > >>>>>>>> > >>>>>>>> > >>>>>> > >>>>>> > >>>>>> > >>>>>> -- > >>>>>> Felix Andrews / 安福立 > >>>>>> http://www.neurofractal.org/felix/ > >>>>> > >>>>> > >>>>> > >>>>> > >>>>> -- > >>>>> Felix Andrews / 安福立 > >>>>> http://www.neurofractal.org/felix/ > >>>>> > >>>>> > >>>>> -- > >>>>> Felix Andrews / 安福立 > >>>>> http://www.neurofractal.org/felix/ > >>>> > >>>> > >>>> > >>>> -- > >>>> Felix Andrews / 安福立 > >>>> http://www.neurofractal.org/felix/ > >>> > >>> > >> > >> > >> > >> -- > >> Felix Andrews / 安福立 > >> http://www.neurofractal.org/felix/ > > > > > > -- Regards Chandan Maruthi
