Machine Learning with Swift

Make design choices

When you have a clear understanding of the goals, constraints, competing solutions, and your data, you can start with defining the technical specifics for your future model. The following questions should be answered before you implement your model:

• Is this a supervised or unsupervised learning problem? Classification or clustering? Discriminative or generative model?
• What is the measurement of success? What is your baseline solution and what are your benchmarks? How do you select the best model? In other words, what is the set of metrics that defines the best model?
• What is your strategy of model quality evaluation? Accuracy, precision-recall, cross-validation, or something else? This depends mostly on what costs more ...