liveBook · Manning

Unlabeled data

As explained in Grokking Machine Learning, Second Edition

Data that does not include labels identifying target values or categories for the problem being considered.

Unlabeled data contains no label identifying the target value or category to predict. Whether a feature counts as a label depends on the problem: an email dataset may be unlabeled if no column matters for the question being asked. section 2.6 section 2.5

Pet images without tags

A dataset containing only cat and dog images is unlabeled when it does not say which animal appears in each image. A model cannot directly predict cat or dog labels, but it can compare images and group similar ones, perhaps separating animals or breeds. section 2.10 section 2.6

Unlabeled data supports several kinds of analysis:

  • Clustering groups similar data points into clusters. section 2.10
  • Dimensionality reduction simplifies data with fewer features while preserving much of its information. section 2.10 section 2.17
  • Generative learning creates new data points resembling the existing dataset. section 2.10
Figure 2.1 Labeled data is data that comes with a tag, or label. That label can be a type or a number. Unlabeled data is data that comes with no tag. The dataset on the left is labeled, and the label is the type of pet (dog/cat). The dataset in the middle is also labeled, and the label is the weight of the pet (in pounds). The dataset on the right is unlabeled.

Compare labeled pet datasets with an image-only dataset to see how the presence or absence of a target changes the learning task. section 2.6

Figure 2.4 An unsupervised learning algorithm can still extract information from data. For example, it can group similar elements together.

See how an unsupervised method can extract structure by grouping similar unlabeled images. section 2.10

How Grokking Machine Learning, Second Edition covers it

How deep it goes
Defines unlabeled data with email and image examples, then names clustering, dimensionality reduction, and generative learning.
Examples
Uses email records without target columns and pet-image collections without animal or weight labels.
What you can do after
Evaluate/understand whether a dataset has a problem-specific target and identify suitable unsupervised uses.
How it connects
Connects unlabeled data to supervised learning, unsupervised learning, clustering, dimensionality reduction, and generative learning.

Also explained in

Compare how 6 books explain unlabeled data →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Grokking Machine Learning, Second Edition. The section links point to the parts of the book it draws on.