Which technique is commonly used in supervised learning?

The technique commonly used in supervised learning involves training a model with labeled datasets. In supervised learning, the learning process operates under the premise that each training example is paired with an output label, allowing the model to learn the relationship between input data and the corresponding outputs. This is fundamental to developing models that can make predictions or classifications based on new, unseen data. The presence of labels provides a clear direction for the learning algorithm to adjust its parameters, optimizing its ability to generalize and make accurate predictions.

In contrast, the other techniques mentioned are more aligned with different learning paradigms. Using data without labels to train refers to unsupervised learning, where the goal is to identify patterns without the guidance of labeled outcomes. Applying algorithms based solely on probabilities is more characteristic of probabilistic models but does not inherently imply the use of supervised techniques. Lastly, finding hidden structures in unlabelled data is another hallmark of unsupervised learning, focusing on extracting insights from datasets lacking explicit labels. Thus, training with labeled datasets is the defining feature of supervised learning, making it the correct choice.