Data science & AI · Reference
What is a support vector machine?
A support vector machine is a supervised machine-learning algorithm that classifies data by finding the boundary that separates the classes with the widest possible margin, and can model non-linear boundaries using kernels.
Maximising the margin
A support vector machine separates two classes by choosing, among the many possible boundaries, the one with the largest margin — the greatest distance to the closest points of either class. The intuition is that a boundary sitting in the middle of the widest gap is the most robust and generalises best. The closest points, which sit on the edge of the margin and determine where the boundary lies, are the support vectors; the rest of the data does not affect the solution, which is part of what makes the method efficient and distinctive.
The kernel trick
Many datasets cannot be separated by a straight line. SVMs handle this with the kernel trick: a kernel function implicitly maps the data into a higher-dimensional space where a linear boundary may exist, without ever computing the coordinates in that space explicitly.
Common kernels include the polynomial and the radial basis function (RBF). This lets an SVM draw flexible, non-linear boundaries in the original feature space while keeping the underlying optimisation tractable.
Soft margins and use
Real data is rarely perfectly separable, so SVMs use a soft margin that allows some misclassifications, governed by a parameter that trades off a wider margin against fewer errors. SVMs were among the most effective classifiers before the rise of deep learning and remain strong on smaller, well-structured datasets, particularly high-dimensional ones such as text. They are less suited to very large datasets, where training cost grows, and to problems where neural networks now dominate.
Support vector machines in research
In research, SVMs are a reliable supervised classifier, especially when data is limited and high-dimensional, where their margin-maximising principle resists overfitting. Results depend on the choice of kernel and its parameters, which are typically tuned by cross-validation; reporting these choices is necessary for reproducibility. As with any classifier, performance should be evaluated on held-out data using appropriate metrics rather than accuracy alone.
Key facts
At a glance
- Definition: classifier maximising the margin between classes
- Type: supervised learning (mainly classification)
- Margin: widest gap between boundary and nearest points
- Support vectors: the closest points that define the boundary
- Kernel trick: enables non-linear boundaries (e.g. RBF kernel)
- Soft margin: allows some misclassification
Common questions
FAQ
How does a support vector machine work?+
An SVM finds the decision boundary that separates the classes with the widest margin — the largest distance to the nearest points of each class. Those nearest points, the support vectors, determine the boundary; the remaining data does not affect it.
What is the kernel trick?+
The kernel trick lets an SVM separate data that is not linearly separable by implicitly mapping it into a higher-dimensional space where a linear boundary exists, without computing that space directly. Kernels such as the radial basis function enable flexible non-linear boundaries.
When are support vector machines a good choice?+
SVMs work well on smaller, high-dimensional datasets such as text, where their margin-maximising approach resists overfitting. They are less suited to very large datasets, where training is costly, and to perceptual tasks now dominated by deep neural networks.
Going deeper
Related on CASRAI
- What is supervised learning? →
- What is machine learning? →
- What is overfitting? →
- What is a neural network? →
- Computer science, data science & AI →
Sources
The step most authors miss
Doing CRediT right? Don’t stop at the statement.
A CRediT statement credits you inside one paper. The recognition CRediT was built for happens when those roles are tied to you, persistently. Sign in with your ORCID — free — and claim your CRediT contributions on casrai.org, the home of the standard. They become a verified, portable part of your identity, not a line that disappears into one PDF.
Free: claim your contributions, then export a journal-ready CRediT statement, schema.org structured data, JATS XML, CSV or BibTeX — and preview your public profile. A membership publishes that profile publicly and verifies the journals you serve.
