Encog 3.0.0 for Java has been released. .Net will follow in a few days.
This release of Encog brings many changes such as multiple RPROP types, QuickProp, PNN/GRNN and less overhead communicating to the flat networks. Major enhancements to the Encog Workbench, as well as the Encog Analyst, which supports automatic normalization.
Downloads are now through GitHub:
Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer
Comments
hello
hello,
Iam still a beginner on this NN stuff. I having some real work in preparing the data for the NN in my project yet (mixing some new source indicators for a kind of financial forecast application...). Soon it will be ready to feed the NN and see the results...
Meanwhile was downloading and testing the encog3 workbench and I have some doubts. I have created a NN with 30 inputs, one hidden layer of 10 and 1 output neuron : NN.eg. Then I made a random csv file with 31 columns (I had to made it in encon2 becouse the min/max values are not set between -1 and 1 as we tell in the dialog on encog3..) then I was able to make the egd by using the convert option on encog3 workbench. then I did the test train then I said yes : save the train. Another file has appeared "NN-cont.eg". So far so good right ? then I made another random file and another train on NN.eg. Then yes save again. I suppose the train of the NN with second file will be a continuation of the first egd file ?
When will the book encog3 (java) came out ? ..so I can see in detail the load and save in code (and also to see what is the significance column (?) :)
BR,
Rui