What Are We Predicting or Grouping?
AI broadly studies systems that perceive, reason, and act toward goals. It includes search, reasoning, planning, and learning. Machine learning learns decision rules from data.
Similar-looking tasks have different goals
- Classification: Predict a predefined category, such as apple versus pear.
- Regression: Predict a numerical quantity.
- Clustering: Group similar records without supplied category labels, then interpret the groups.
A classification target is a label. Learning from labels is supervised learning; finding structure without supplied answers, as in clustering, is unsupervised learning.
Must clustering come first?
No. Human-labeled fruit records can train a classifier directly. Clusters also need not match the categories we want. Grouping by size might combine large apples and small pears.
“Divide the data well” is too vague to evaluate. Specify what to predict, what errors cost, and how the result will be used.
Check your understanding
How do predicting a new song’s genre from labeled examples and grouping listeners by similar habits differ?
Show explanation
The first is classification using genre labels. The second is clustering if no predefined listener-type answers are supplied. Giving a group a name afterward does not itself make the learning supervised.