11 Practical Methodology

11.1 Performance Metrics

11.2 Default Baseline Models

11.3 Determining Whether to Gather More Data

11.4 Selecting Hyperparameters

11.4.1 Manual Hyperparameter Tuning

Effective capacity is constrained by three factors: the representational capacity of the model, the ability of the learning algorithm to successfully minimize the cost function used to train the model, and the degree to which the cost function and training procedure regularize the model.

11.4.2 Automatic Hyperparameter Optimization Algorithms

11.4.3 Grid Search

11.4.4 Random Search

11.4.5 Model-Based Hyperparameter Optimization

11.5 Debugging Strategies

11.6 Example: Multi-Digit Number Recognition