Tuesday, April 20, 2010

2nd public commercial HTM application

Looks like a press release was issued today by EDSA Power Analytics, noting that it is teaming up with Numenta to develop autonomous monitoring of electrical power systems. Apparently, for some applications, electrical power failures are hugely expensive. EDSA develops software to monitor power systems to attempt to prevent power failures. HTM can be used by EDSA to learn the difference between normal and non-normal electrical activity, such that the HTM becomes increasingly able to predict a future power failure.



  1. Great - thanks for the link!

    Unrelated question: Have you heard anything about, or have any ideas on, Numenta integrating memristor technology into their work?


  2. Memristor ? They were made in 2008 by HP, and it's still a research project. I can't imagine Numenta doing anything with memristors at this point.

  3. Maybe not right away, but within 3 years they'll be readily available:


    In any case, I thought they might have some ideas about how to potentially integrate them.
    Here's more:



  4. Dave:

    I actually read that article that you are quoting. Certainly it appears that memristors are getting closer to commercialization. I haven't heard anything about Numenta doing that, but in the last year or so (see Hawkins "Grand Challenges" speech from last March 2009 and Dileep George's June/July 2009 paper) it is becoming clear that Numenta is starting to think seriously about implementing its algorithms in hardware.

    Actually, before Hawkins' recent speech on the new learning algorithms, I would have said that HTM appears to be emulating the brain at too high of a level for synapses to come into play. Now, however, it looks like the new, more biologically realistic algorithms are emulating the functions of at least some synapses in the neuron. What isn't clear to me is whether that emulation is merely "on-off" synapses or whether the more realistic memristor synapses that can have a large number of different values resembles what Numenta is doing. Hopefully Numenta does put out a white paper on the new algorithms to fill in the details.

  5. Thanks for the response, Sean. I'd think memristors are enough of a game changer that they'd even think about modifying their algorithms to incorporate them. It seems like any realistic model of the cortex would have the incorporate synaptic weights, so I'd be surprised if they couldn't make use of them.

    Definitely let us know if you hear anything. Thanks again!

  6. Definitely will. And I am sure that Numenta will have its eye on any hardware that will make it easier to emulate the cortex.

    I hope they look into running their software on GPU's. From what I have seen, GPU's could potentially speed up their software by hundreds of times.

  7. Here is a response from Numenta on the subject:


  8. The HTM approach based memristors for both memory and transistor replacement is the ticket. The big shebang. The end of biological brains as something useful in the marketplace.