Artificial Intelligence By Example

Measuring the precision of prediction of a network through accuracy values

The accuracy node contains two parts: correct prediction and accuracy. The output value is not sent to the training node but the correct predictions (see next section, Correct prediction) part of it is used by cross-entropy. Accuracy measures the performance of the system. You can stop training by limiting the number of epochs hyperparameter if you estimate that the accuracy is sufficient to achieve your goals. Or you can use accuracy to fine-tune the training model to increase its rate faster by modifying the hyperparameters or the design of your model.