Natural Language Processing with Python Quick Start Guide

Loss function

One of the pieces that's used to set up a machine learning task is to assess how a model is doing. The simplest answer would be to measure the notional accuracy of the model. Accuracy has few flaws, though:

• Accuracy is a proxy metric tied to validation data and not training data.
• Accuracy measures how correct we are. During training, we want to measure how far our model predicts from the target.

These differences mean that we need a different function to meet our preceding criteria. This is fulfilled by the loss function in the context of deep learning. This is sometimes referred to as an objective function as well.

"The loss function takes the predictions of the network and the true target (what you wanted the network to ...