All geek questions in one place

Adding a Neural Displacement System to the Encog Network

I am porting some C code including an artificial neural network in Java. I decided to use the Encog framework .

I would like my network to have a neural bias system (which uses weighted synapses and always outputs 1.0) connected to hidden layers and the output layer; they are marked in green in this graph .

How can I create such a network programmatically? Should each of the neural neurons have its own layer? I tried to create an instance of BasicLayer for each neural syndrome, but this does not seem to have any effect, even if the layer is connected through WeightedSynapse.

Thanks,

Martin Vibo

+4
source share
1 answer

Ok, I know that! I am one of the main developers of Encog. :)

We have improved support for this in Encog 2.4 (which is in beta), however even the current version supports this. In Encog 2.3, the offset is called a threshold. Therefore, create your layers by indicating that you WANT threshold values. This basically creates bias inputs for each of the levels you specify having threshold values. I build it all in one layer, because it improves productivity, everything works the same way.

We really discussed this, and we clarify it in Encog 2.4. You can read the discussion here.

http://www.heatonresearch.com/node/1477

+4
source

Source: https://habr.com/ru/post/1311794/

More articles:


All Articles