Which type of algorithms typically has a fixed number of parameters?

The correct choice is parametric algorithms. These types of algorithms are characterized by their fixed number of parameters, which are determined prior to the model training process. For instance, in linear regression, the relationship between the input variables and the output variable is expressed through a fixed number of parameters, such as the slope and intercept. Once these parameters are set, the learning algorithm aims to estimate their values using the training data.

In contrast, supervised learning algorithms may include both parametric and non-parametric algorithms, which means they can either have a fixed number of parameters or adapt their complexity based on the data. Hyperparameter tuning algorithms focus on optimizing the hyperparameters of a model rather than being characterized by a fixed number of parameters themselves. Reinforcement learning algorithms operate within a different paradigm, learning optimal actions through interactions with an environment and often involving a more dynamic set of parameters that change over time as the model learns from experiences. Thus, the essential defining feature of parametric algorithms is their fixed number of parameters, making this choice the most accurate.