Dictionary <string, MMNode> createModels(List <MMNodeFactory.Model> modelsToTrain)
{
Dictionary <string, MMNode> models = new Dictionary <string, MMNode>();
int neuronsCount = 50;
//for (int neuronsCount = 10; neuronsCount < 100; neuronsCount += 20)
{
foreach (MMNodeFactory.Model selectedModel in modelsToTrain)
{
//Create the model
//MNNodeFactory.Model selectedModel;
//Enum.TryParse<MNNodeFactory.Model>(uncastedMdl.ToString(), out selectedModel);
models[selectedModel.ToString() + "_" + neuronsCount] = MMNodeFactory.obtain(selectedModel, neuronsCount);
MMNode network = models[selectedModel.ToString() + "_" + neuronsCount];
network.onEpoch += network_onEpoch;
//network.addModality( new Signal(2,1), "XY-t0");
//network.addModality( new Signal(2,1), "XY-t1");
network.addModality(new Signal(retinaSize * 4, retinaSize), "Vision-t0-Color");
network.addModality(new Signal(retinaSize * 2, retinaSize), "Vision-t0-Orientation");
//network.addModality(new Signal(retinaSize * 4, retinaSize), "Vision-t0-Shape");
network.addModality(new Signal(4, 1), "Saccade");
network.addModality(new Signal(retinaSize * 4, retinaSize), "Vision-t1-Color");
network.addModality(new Signal(retinaSize * 2, retinaSize), "Vision-t1-Orientation");
//network.addModality(new Signal(retinaSize * 4, retinaSize), "Vision-t1-Shape");
//Apply a treshold function on the modalities
network.onDivergence += network_onDivergence;
}
}
return(models);
}
public HippocampusForm()
{
InitializeComponent();
//--------------------------------------------------------------------------
//-------------MEC
mec = new List <IONode>();
int mecAreas = 1;
for (int i = 1; i <= mecAreas; i++)
{
mec.Add(
new IONodeGridCells
(
new Point2D(2, 1), //Input, we cheat, this is the odometer values X, Y
new Point2D(10, 10), //Number of grid cells This defines the granularity, it should vary among the MEC
0.1, //rf size
0.5 * i //spacing
));
flowLayoutPanelMEC.Controls.Add(mec.Last().GetCtrl());
}
//--------------------------------------------------------------------------
//--------------LEC
lec = new List <IONode>();
int lecAreas = 1;
for (int i = 1; i <= lecAreas; i++)
{
lec.Add(
new IONodeAdaptiveSOM
(
new Point2D(foveaSize, foveaSize), //Input, size of the fovea
new Point2D(20, 20), //Size of the map. Defines the number of templates/filter used.
true) //USe only winner as output
);
flowLayoutPanelLEC.Controls.Add(lec.Last().GetCtrl());
}
//--------------------------------------------------------------------------
//--------------CA3
CA3Inputs = new List <SignalLink>();
CA3 = new MMNodeSOM
(
new Point2D(20, 20), //Size of the map.
true //Use only the winner for prediction
);
//Add all the MEC modalities
int counter = 0;
foreach (IONode n in mec)
{
SignalLink link = new SignalLink(n.output, new Signal(n.output));
CA3.addModality(link.to, "MEC_" + counter++); //note: n.output is cloned, not a reference
CA3Inputs.Add(link);
}
//Add the LEC modalities
counter = 0;
foreach (IONode n in lec)
{
SignalLink link = new SignalLink(n.output, new Signal(n.output));
CA3.addModality(link.to, "LEC_" + counter++); //note: n.output is cloned, not a reference
CA3Inputs.Add(link);
}
ctrlMMNode1.attach(CA3);
//--------------------------------------------------------------------------
//--------------CA1
CA1 = new MMNodeLookupTable
(
new Point2D(1, 1) //Size of the map.
);
CA1.addModality(new Signal(foveaSize, foveaSize), "FOVEA");
CA1.addModality(new Signal(2, 1), "ODOMETRY");
ctrlMMNode2.attach(CA1);
}