Feature union in ML Python - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is a Feature Union in machine learning?

Feature Union is a technique that combines multiple feature extraction processes into one. It joins different sets of features side by side to create a bigger feature set for a model.

beginner

Why use Feature Union instead of just one feature extractor?

Using Feature Union lets you combine different types of features that capture different information. This can help the model learn better by having more diverse data.

intermediate

How does Feature Union work internally?

Feature Union runs each feature extractor separately on the input data, then stacks their outputs horizontally (side by side) to form one combined feature matrix.

intermediate

Give a simple example of using Feature Union in Python with scikit-learn.

You can combine a text vectorizer and a numeric scaler like this: FeatureUnion([('text', CountVectorizer()), ('num', StandardScaler())]). This creates features from text and numbers together.

intermediate

What is the difference between Feature Union and Pipeline in scikit-learn?

Feature Union combines features from parallel transformers side by side. Pipeline applies transformers sequentially, one after another, transforming data step by step.

What does Feature Union do with the outputs of multiple feature extractors?

A Stacks them side by side to create a combined feature set

B Adds them together element-wise

C Selects the best output and discards others

D Runs them one after another in sequence

Feature Union stacks outputs horizontally, combining features side by side.

Which scikit-learn class is used to perform Feature Union?

A FeatureUnion

B Pipeline

C GridSearchCV

D StandardScaler

FeatureUnion is the class designed to combine multiple feature extractors.

Feature Union is most useful when:

A You want to apply transformations sequentially

B You want to reduce the number of features

C You want to combine different types of features from the same data

D You want to train multiple models separately

Feature Union helps combine different feature sets extracted in parallel.

What is the main difference between Feature Union and Pipeline?

A Feature Union is for model training; Pipeline is for data cleaning

B Feature Union applies steps sequentially; Pipeline combines features side by side

C Feature Union is only for numeric data; Pipeline is only for text data

D Feature Union combines features side by side; Pipeline applies steps one after another

Feature Union merges features horizontally; Pipeline chains transformations vertically.

If you want to combine text features and numeric features for a model, which technique is best?

A Grid Search

B Feature Union

C Cross Validation

D Principal Component Analysis

Feature Union allows combining different feature types into one feature set.

Explain in your own words what Feature Union is and why it is useful in machine learning.

Think about how combining different views of data can help a model.

Describe the difference between Feature Union and Pipeline in scikit-learn.

Consider how data flows through each.