Strong learner

As explained in Grokking Machine Learning

A strong learner is an ensemble model formed by combining multiple weak learners to achieve substantially improved predictive performance.

Imagine asking several friends to answer an exam. A majority vote can be more reliable than trusting one person, especially when each friend is good at different questions. A strong learner uses the same idea: it combines multiple weak learners into one model with substantially better predictive performance. section 12.1

The combination depends on the task. Classifiers vote for the most common class, while regressors average their predictions. The weak learners are often decision trees, whose simple boundaries become more useful when combined. section 12.1 section 12.2

Figure 12.1 Ensemble methods consist of joining several weak learners to build a strong learner.

See how several weak learners join to form a strong learner. section 12.1

Two main ensemble strategies build strong learners in different ways:

• Bagging trains learners on random dataset subsets, then combines their predictions by voting or averaging. Random forests are a worked example. section 12.1 section 12.2
• Boosting trains learners sequentially, enlarging or emphasizing poorly predicted points so later learners focus on earlier weaknesses. section 12.1 section 12.3
• AdaBoost gives learners voting power based on accuracy scores, whereas gradient boosting fits successive learners to residual errors. section 12.3 section 12.4

Spam classification with a random forest

For spam and ham emails represented by word counts, three depth-1 decision trees can be trained on random subsets. Each predicts spam or ham; the strong learner chooses the most common prediction. The resulting boundary classifies most points while treating isolated mistakes as outliers rather than overfitting them. section 12.2

A boosted regression prediction

For user age and weekly app engagement, gradient boosting starts with the average engagement, fits a tree to the residuals, and adds later tree predictions using a learning rate. As residuals shrink, the later weak learners make smaller corrections. section 12.4

You can practice the idea by combining three weighted classifier predictions and determining the final label, then calculating an AdaBoost learner's weight from its accuracy. section 12.8

How Grokking Machine Learning covers it

How deep it goes

Explains voting, averaging, reweighting, residual fitting, and manual tree ensembles, while deferring detailed XGBoost implementation differences.

Examples

Uses friend voting, spam email classification, random forests, AdaBoost diagrams, user engagement regression, and weighted-vote exercises.

What it assumes

Assumes basic decision-tree familiarity while defining learner as another name for a machine-learning model.

What you can do after

Apply it via Scikit-Learn ensemble classes and reason through voting, learner weights, residuals, and learning-rate updates.

How it connects

Connects weak learners with ensemble learning, bagging, boosting, random forests, AdaBoost, gradient boosting, XGBoost, and decision trees.

liveBookmanning.com

Read Grokking Machine Learning on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Ensemble Methods for Machine LearningA strong learner is a sufficiently complex predictive model with good performance on its task.

Compare how 2 books explain strong learner →

Related concepts

• boostingproduced by · Boosting is an ensemble-learning technique that builds predictive models sequentially and combines them, with later models improving on weaknesses of earlier ones.
• sequential ensemble methodproduced by · A sequential ensemble method trains base learners in successive stages, with later learners depending on or correcting the outputs of earlier learners.
• weak learnerhas part · A weak learner is a predictive model that performs only modestly better than chance or a simple baseline and can contribute to a stronger ensemble.
• parallel ensemble methodused by · A parallel ensemble method combines outputs from multiple independently trained base learners that perform the same task.
• AdaBoostproduced by · AdaBoost is an ensemble-learning algorithm that sequentially combines weak learners, adaptively reweighting training examples so that later learners emphasize those earlier learners handled poorly.
• random forestincludes · A random forest is an ensemble model that combines predictions from multiple decision trees built with randomized data samples, feature selections, or both.

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