On Thu, Jun 30, 2022 at 1:33 PM Boris Kazachenko <cogno...@gmail.com> wrote:
> On Thursday, June 30, 2022, at 3:00 AM, Rob Freeman wrote: > > I'm interested to hear what other mechanisms people might come up with to > replace back-prop, and do this on the fly.. > > > For shared predictions, I don't see much of an alternative to backprop, it > would have to be feedback-driven anyway. Which means coarse and wasteful. > > My alternative is to directly search for shared properties: lateral > cross-comparison and connectivity clustering. It's a lot more complex, but > that complexity is intelligent, it's not just throwing more brute force on > the same dumb algorithm. > Well, I agree with you on the second part: "directly search for shared properties: lateral cross-comparison and connectivity clustering". But I don't see why that could not be used to group for shared predictions. If you made prediction the "shared property", doing a "connectivity clustering" of shared predictions, which is to say making "prediction" or sequence the connectivity, then that would group for shared predictions, wouldn't it? ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T5d6fde768988cb74-M84a50c3756ea7ab8b62ca816 Delivery options: https://agi.topicbox.com/groups/agi/subscription