In the previous chapter, the ability to use the DL4J UI to monitor and debug a Multilayer Neural Network (MNN) was fully described. The last part of the previous chapter also explained how to interpret and use the real-time visual results in the UI charts to tune training. In this chapter, we will explain how to evaluate the accuracy of a model after its training and before it is moved to production. Several evaluation strategies exist for neural networks. This chapter covers the principal ones and all their implementations, which are provided by the DL4J API.
While describing the different evaluation techniques, I have tried to reduce the usage of math and formulas as much as possible and keep the focus ...
