Get Started

Sharp Brain's Interactive Editor creates a named pipe "SharpBrain", All the interaction of Neural Network and your external application are made through that pipe only by sending a request string.

Request is defined as space separated key-value pair or flag. example: "key1=value1 flag1"

Network finds suitable neuron with the same key and feed it with a value as defined in request package. Hence Neural Network can be made forward propagation through a series of key value pair which covers all input neurons (parent.count == 0) Training is done by sending expected output of output neuron through same key-value pair.

Flags

• -e Request Network to output error on pipe client buffer. Generally it is used only if back propagation is performed, But if it is not so then it will output 0.

Note Network always assumes that you either feeding output (training) or requesting output of every terminal neuron. If you missed any terminal neuron in training request package, Network will print output of all terminal neurons which are not present in requested package

##Example Code

var client = new NamedPipeClientStream("SharpBrain");
client.Connect();

var reader = new StreamReader(client);
var writer = new StreamWriter(client)
{
     AutoFlush = true
};

writer.WriteLine("n0=0 n1=1 o1=5");//feed 0 1 and train against 5
writer.WriteLine("n0=0 n1=1 o1=5 -e");//feed 0 1 and print output error
writer.WriteLine("n0=0 n1=1");//print output of neuron o1