Support Vector Machines

Flashcards covering key concepts and terminology related to Support Vector Machines (SVM) in Machine Learning.

Support Vector Machine (SVM)

A powerful Machine Learning model used for classification, regression, and outlier detection.

Large Margin Classification

A method where the decision boundary of an SVM classifier is as far away from the closest training instances as possible.

Support Vectors

Instances located on the edge of the margin that fully support the decision boundary.

Hard Margin Classification

A strict approach that requires all instances to be off the street and on the correct side, only applicable to linearly separable data.

Soft Margin Classification

A more flexible approach that allows for some margin violations to balance the width of the street and accuracy.

Hyperparameter C

Controls the balance between maximizing the margin and minimizing margin violations in SVM.

Kernel Trick

A mathematical technique that allows SVM to work in high-dimensional space without explicitly adding more features.

Polynomial Kernel

A kernel that allows SVM to model complex relationships by using polynomial features.

Gaussian Radial Basis Function (RBF) Kernel

A similarity function used in SVM to transform input data into a higher-dimensional space.

Epsilon-Insensitive Loss

In SVM regression, it defines a width around the prediction line where errors do not penalize the model.

Scikit-Learn

A popular Python library used for machine learning, providing tools for implementing SVM and other algorithms.

Linear SVC

A Scikit-Learn class used for linear support vector classification.

SVR

Support Vector Regression class in Scikit-Learn that uses the kernel trick for regression tasks.

Margin Violations

Instances that fall inside the margin, which are allowed in soft margin classification to improve model generalization.

Feature Scaling

Standardizing features to ensure that the features contribute equally to the distance computation in SVM.

Grid Search

A method to tune hyperparameters by exhaustively searching over specified parameter values.