FENNIX

Fast Experimentation with Neural Networks


Code examples

This section shows some useful examples of FENNIX scripts performing different tasks:

Reading a dataset


	data = call(FNNXDatasetLoader) {
		file = 	"dataset_file_name.fxd"
		format = 	"FENNIX_008"
	}
	
Back to top

Normalizing a dataset


	call(FNNXNormalizer) {
		dataset = 	data
		method = 	"MIN_MAX"
		p1 = 	-0.95
		p2 = 	0.95
	}
	
Back to top

Creating a multi-layer perceptron


	network = new(FNNXMLP) {
		inputs = 	3
		hidden = 	2
		actFuncHidden = 	"TANH"
		outputs = 	1
		actFuncOutput = 	"TANH"
		shortcut = 	false
	}
	
Back to top

Creating a plotter


	plot = new(FNNXPlotter)
	
Back to top

Creating a saver


	save = new(FNNXSaver) {
		file = 	"results_file_name"
	}
	
Back to top

Training a multi-layer perceptron using the backpropagation algorithm


	test = new(FNNXSSETester) {
		network = 	network
		dataset = 	data
	}
	new(FNNXOnlineBackprop) {
		network = 	network
		dataset = 	data
		tester = 	test
		etaIni = 	0.05
		etaEnd = 	0.01
		mu = 	0.7
		epochs = 	100
		weightDecay = 	0.0
		minErrorPercent = 	2
		stopIfTesterBelow = 	0.0
		logCounter = 	100
		delay = 	0
		plotter = 	plot
		saver = 	save
	}
	
Back to top

Computing the outputs of a multilayer perceptron


	call(FNNXOutputTester) {
		network = 	network
		dataset = 	data
		saver = 	save
		plotter = 	plot
	}
	
Back to top

Testing the performance of a multilayer perceptron


	call(FNNXPerformanceTester) {
		network = 	network
		dataset = 	data
		saver = 	save
		plotter = 	plot
		allowedError = 	40
	}
	
Back to top

Exporting a multi-layer perceptron


	call(FNNXExportMLP) {
		network = 	network
		saver = 	save
	}
	
Back to top

Sampling a dataset


	dataSampled = call(FNNXSampler) {
		dataset = 	data
		percent = 	80
	}
	
Back to top

Last modification 10 April 2012
by Héctor Satizábal.