Which type of algorithms are characterized by generating a potentially infinite number of model parameters?

Non-parametric algorithms are designed in such a way that they do not operate under the assumption of a fixed number of parameters. Instead, they adapt and grow in complexity according to the amount of data they are trained on. This characteristic allows them to generate potentially infinite model parameters as more data points are introduced. Non-parametric methods can flexibly adjust to the underlying data distribution, making them particularly useful for complex problems where the structure of the data is not known in advance.

In contrast, parametric algorithms assume a specific form for the underlying data distribution, typically characterized by a fixed number of parameters. Consequently, these algorithms do not generate an infinite range of parameters but rather depend on the pre-defined structure.

Supervised and unsupervised algorithms refer to the nature of the learning process rather than the parameterization. Supervised algorithms learn from labeled data, while unsupervised algorithms work with unlabelled data to find inherent patterns. Both types can utilize either parametric or non-parametric methods, but their definitions do not pertain to the infinite parameter generation characteristic.