Evaluating models
Model evaluation in AI and machine learning (ML) models typically relies on the methodology of selecting training, validation, and test sets and loss functions and metrics that are calculated on those datasets. This enables us to have a quantitative way to assess models and optimize them. For instance, we need a way to measure how good our malware detection model is in classifying executables during the training time, validation time, and test time.
Firstly, we’ll describe and compare the most common loss functions used in different types of ML models.
Loss functions
Loss functions are used when training the model to evaluate its performance during training. Furthermore, they are used to compute gradients, which enable us to determine how to change model parameters to gradually improve the performance of the model in each training step. Loss functions can also be used to evaluate models after training.
The following loss functions are the most used...
