CODEWITHSUNDEEP
neural-networkencog
Arsen
Arsen

Reputation: 143

Encog getError() returns Infinity

I built following neural network with help of Encog library for Java

network.addLayer(new BasicLayer(DataCooker.DATA_SIZE));
network.addLayer(new BasicLayer(DataCooker.DATA_SIZE));
network.addLayer(new BasicLayer(DataCooker.DATA_SIZE));
network.addLayer(new BasicLayer(DataCooker.DATA_SIZE));
network.addLayer(new BasicLayer(1));
network.getStructure().finalizeStructure();
network.reset();

Also, I prepared test-data and tried to train this NN with help of this train

Train train = new ResilientPropagation(network, trainingSet);

But, i found that on some data case (rare) I am receiving train.getError() equals to Infinity (not depends how many epochs passed). Data looks good from first glance (all are numbers, no NaN, no Infinity values).

What is possible reasons for this Infinity error? What can I do to solve it ?

Thanks

Upvotes: 1

Views: 332

Answers (1)

Arsen
Arsen

Reputation: 143

sorry, data was bad indeed, output for bad cases was Infinity

Upvotes: 1

Related Questions