Which term refers to the practice of giving different weights or importance to different components within a model?

The term that refers to the practice of giving different weights or importance to different components within a model is indeed the concept of weighted learning. This approach allows for the assignment of varying levels of influence to different features or observations based on their relevance or importance in contributing to the model's predictive accuracy.

Weighted learning is particularly useful in scenarios where certain data points should impact the model's learning process more than others due to their specific significance, reliability, or prevalence. For instance, in a classification task, misclassifying instances from a minority class might be more critical than misclassifying those from a majority class; thus, heavier weights can be applied to minority instances to ensure the model focuses on minimizing those errors.

In contrast, the other concepts—such as weight adjustment, regularization, and parameter tuning—address different aspects of model optimization and training. Weight adjustment could imply changing weights numerically without the broader context of the learning strategy. Regularization refers to techniques aimed at reducing overfitting by penalizing large coefficients in regression models. Parameter tuning involves altering model parameters to achieve optimal performance but does not specifically encompass the notion of assigning different importances to various components as weighted learning does.