Finding the optimal model parameters with HyperDrive
In ML, we typically deal with either parametric or non-parametric models. Models represent the distribution of the training data to make predictions for unseen data from the same distribution. While parametric models (such as linear regression, logistic regression, and neural networks) represent the training data distribution by using a learned set of parameters, non-parametric models describe the training data distribution through other traits, such as decision trees (all tree-based classifiers), training samples (k-nearest neighbors), or weighted training samples (support vector machine).
Parametric models such as linear or logistic regression are typically defined by a constant number of parameters that are independent of the training data. These models make strong assumptions about the training data, so they often require fewer training samples. As a result, both training and inferencing are usually very fast.
In comparison...