Bernard

I discovered what the issue is: The formula in the client spreadsheet ony works after the sheet was saved at lease once. It does not work in a new unsaved sheet for some reason. It is not a server issue. My apologies.

Thanks for providing this well-documented tool! How can I request from within an Excel sheet sample data from the running MyServer? I tried entering the following formula in a cell: =DDE("MyServer", "MyAdvise", "MyTopic") But this fails with "#N/A" in the cell, from both Excel and LibreOffice Calc. Many thanks.

Thanks Zoran. I made a change converting from List to array but that confirms what you wrote. Surprisingly this did not change the profiling result, so it appears on the surface that the Hotspot JVM is very smart. BTW my application trains the network in mutiple threads created by a Watchmaker GA, and I found this to run much faster than the same setup with DL4J on a 4 core CPU (NO GPU). IMHO Neuroph is an excellent framework for serious experimentation, and multi-treading on a multi core CPU goes...

Hi, I noticed that Neuroph performance is quite good when I run it in multiple threads. Still I would want it to be faster. After profiling my case, I found that the prerformance hotspot is in org.neuroph.core.input.WeightedSum.getOutput(List<Connection> inputConnections) where it iterates it over a List. This is very slow. Are there any plans to use an array to speed up proccessing?

For paralleelism there is an answer here: Help with Neuroph neural network. MultiLayerPerceptron#learnInNewThread()

Because Neuroph must add 1 BiasNeuron to the hidden layer, same as q to the input layer which has 14 + 1 = 15 neurons