Hey Jim, Yes, it absolutely crossed my mind to post this on the encog-java forum. So I did just that (see here <http://www.heatonresearch.com/node/2716>). Thanks very much for your quick responses so far.
Tim Washington Interruptsoftware.ca 416.843.9060 On Sun, Aug 5, 2012 at 3:04 PM, Jim - FooBar(); <jimpil1...@gmail.com>wrote: > Hi Tim, > > first of all thanks for opening a new thread for this... > > 2 things : > > The 1st paragraph is really a perfect question for the encog-java > forum...I'm not sure anyone here can help with this (not even me - > financial prediction with temporal data is one of those encog corners I've > not had the chance to look at or experiment yet)...I do believe however > that posting such a well phrased and genuinely good question will attract > attention, perhaps from Jeff Heaton himself who you can think of as > 'reliable'...just don't expect things to go as quickly as they are going > here on the clojure ml... > > Now, on the 2nd part of your question which is actually about > clojure-encog... > > > Back over the clojure-encog, the thing that normalizes input data, the > make-data<https://github.com/jimpil/clojure-encog/blob/master/src/clojure_encog/training.clj#L41>function > only deals with doubles (not a list of tick data entries). > > > the temporalDataset and foldedDataset (perhaps others too) have not been > wrapped yet...Patches are welcome if you get it going finally... :). The > most common ones (array-1d or array-2d) are there though. > > The make-trainer and > train<https://github.com/jimpil/clojure-encog/blob/master/src/clojure_encog/training.clj#L137>functions > seem to iterate for the number of strategies that you've > specified. > > > interrupting iterations can be done via a couple of ways...generally, > either you've got an iteration limit or some error-tolerance that you aim > for. strategies are completely optional and have nothing to do with > signalling end of iterations...in fact, the XOR example works better > without any strategies whatsoever! > > > > But I can't see in > BackPropogation<https://github.com/encog/encog-java-core/blob/master/src/main/java/org/encog/neural/networks/training/propagation/back/Backpropagation.java>or > it's > superclass<https://github.com/encog/encog-java-core/blob/master/src/main/java/org/encog/neural/networks/training/propagation/Propagation.java>, > where that tick data is actually processed (init and iteration methods seem > to just setup a background process). So I'm left wondering how I can give > the core encog neural-net a list of tick data that has a second or > sub-second granularity? > > > I think you're looking at the wrong place...the BackPropagation class > simply defines the algorithm - I don't think you will find anything related > with tick-data there...my bet would be the temporalDataset class which is > actually quite massive and scary! again though, this is a perfectly valid > question for the official encog forum...You can expect Jeff or Seemagh to > respond within the week. I do remember some other guy very much interested > in financial predictions but he was doing everything in C#... > > > Hope that helps, > > Jim > > -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
