Semi-supervised learning

As explained in Causal AI

Semi-supervision is a learning setting that combines labeled and unlabeled examples to train a model.

Semi-supervised learning combines labeled pairs (X, Y) with unlabeled feature examples (X). Labeled data supports learning P(Y|X), while unlabeled data informs P(X); the practical question is whether that extra information improves prediction. section 5.3.2

The approach is most useful when labels are costly but examples are plentiful. A causal interpretation asks whether unlabeled X updates the joint representation P(X, Y) in a way that improves prediction of Y. section 5.3.2

Causal structure supplies a test for when unlabeled examples should help:

• Expect stronger justification in anti-causal learning, where the label causes the features. section 5.4
• Expect weaker justification in causal learning, where the features cause the label. section 5.4
• Use these distinctions to avoid spending labeling effort, computation, and personnel time on unsuitable problem settings. section 5.4

Image moderation

A social media platform may have many uploaded images but few labels indicating gratuitous violence. Human labeling is expensive and mentally stressful; semi-supervised learning aims to reduce the required labeling workload. section 5.3.2

Figure 5.16 In causal learning, the features cause the label. In anti-causal learning, the label causes the features.

Compare causal and anti-causal directions to assess whether unlabeled feature data is likely to inform label prediction. section 5.4

How Causal AI covers it

How deep it goes

Explains data roles, distributional motivation, and causal success conditions without detailing a specific training algorithm.

Examples

Uses unlabeled digit images in a VAE-based causal model and violence-image moderation with costly human labels.

What it assumes

Assumes familiarity with features, labels, conditional distributions, and causal interpretations of variables.

What you can do after

Evaluate whether unlabeled examples are likely to improve prediction by examining causal direction and labeling costs.

How it connects

Connects supervised learning, unsupervised learning, causal learning, anti-causal learning, deep learning, and variational autoencoders.

liveBookmanning.com

Read Causal AI on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Outlier Detection in PythonSemisupervised learning is a machine-learning paradigm that trains with labeled data while also exploiting substantially larger amounts of unlabeled data.
• 

  liveBookmanning.com
  Financial AI in PracticeSemi-supervised learning uses a limited amount of labeled data together with a larger pool of unlabeled data.
• 

  liveBookmanning.com
  Graph Neural Networks in ActionSemi-supervised learning is a learning setting that combines labeled and unlabeled examples during model development.
• 

  liveBookmanning.com
  Knowledge Graphs and LLMs in ActionSemi-supervised learning is a machine-learning paradigm that learns from a combination of labeled and unlabeled examples.

Compare how 5 books explain semi-supervised learning →

Related concepts

• unsupervised learninguses · Unsupervised learning is a machine-learning approach that discovers patterns, structure, or representations in data without labeled target outputs.
• supervised learninguses · Supervised learning is a machine-learning approach that learns to map inputs to target outputs from examples paired with known labels or values.
• labeled dataused by · Labeled data is a set of examples or observations paired with target values, categories, or other task-relevant annotations.
• node classificationused by · Node classification is a graph-learning task that predicts one or more discrete category labels for individual nodes in a graph.
• link predictionused by · Link prediction is a graph-learning task that estimates whether a connection exists or may occur between nodes, often by scoring candidate links.

This explainer was generated by Manning's knowledge graph from the text of Causal AI. The section links point to the parts of the book it draws on.