The risk of an optimal model describes an empirical method for determining the optimal model for a given problem. It accomplishes this task by evaluating all candidate models from the hypothesis space on a sampled dataset and selecting the model with the minimum risk.
The symbol denotes the set of possible models, often from a particular class like "polynomials of any degree" or "multi-layer perceptron networks". For any learning algorithm, indicates the space where an optimal model may be found.
The symbol denotes the optimal model for a problem. It yields the lowest risk for pairs of inputs and outputs. The goal of machine learning is to optimize until it becomes .
Suppose, we have the following models with their empirical risk calculated on an arbitrary dataset of samples:
Using the equation described above, we conclude observe that the optimal model is the model h with the lowest risk.
Therefore, we obtain = .