Skip to content

Support Vector Machines

I. Finding the optimal boundary via margin maximization — overview of SVM

    %%{init: { 'theme': 'base', 'themeVariables': { 'edgeLabelBackground': '#fff' }}}%%
flowchart LR
    A1["Data that is hard to separate linearly"] -- "Kernel trick and margin maximization" --> B1["Robust decision boundary"]
    style A1 fill:#f9f9f9,stroke:#333,stroke-width:1px
    style B1 fill:#e1f5fe,stroke:#01579b,stroke-width:1px
  

Definition: a supervised learning algorithm that performs classification and regression by finding the optimal hyperplane ( Hyperplane ) that maximizes the margin ( Margin ), the distance between two classes, in the space where the data resides

Characteristics: ( Maximum Margin ) maximizes the buffer space between the decision boundary and the data, securing strong generalization performance on unseen data ( Support Vectors ) performs efficient computation using only the key data points ( Support Vectors ) that contribute to forming the decision boundary, rather than the entire dataset ( Kernel Trick ) a technique that maps data that cannot be linearly separated in a low-dimensional space into a higher dimension to create a non-linear boundary

II. Detailed mechanisms and components of SVM

A. The classification mechanism of SVM

    graph LR
    A2["Class A"] <-->|"Margin"| B2["Decision Boundary\n(Hyperplane)"]
    B2 <-->|"Margin"| C2["Class B"]
  

B. Core components and detailed functions

ComponentDetailed DescriptionNotes
HyperplaneThe optimal N-dimensional plane that separates data into different classesDecision Boundary
MarginThe perpendicular distance between the support vectors and the decision boundary — an indicator of model robustnessDistance
Kernel FunctionTransforms non-linear data into a higher-dimensional space via functions such as RBF and PolynomialKernel Function
Slack VariableA flexibility parameter that allows some misclassification in order to prevent overfittingSoft Margin (C)

III. Key characteristics and technology trends of SVM

A. Advantages and limitations

ItemDetailed ContentNotes
Key AdvantageEffective even in high-dimensional spaces and robust against overfittingRobustness
LimitationComputational complexity increases on large datasets, and choosing an appropriate kernel can be difficultComplexity
ApplicationsHigh-dimensional data domains such as text classification, image recognition, and biometric analysisApplication

B. Technology trends

( Baseline Model ) it is still used as a powerful classifier that can substitute for deep learning in settings with limited data and many features. ( Hybrid Approach ) it is increasingly used as the classifier in the final layer of a neural network, or combined with hyperparameter optimization techniques to boost performance.