Support Vector Machines (SVMs)

Finds hyperplane in n-dimensional space that divides the data into the desired classes

Specifically finds the hyperplane with the maximal margin (maximum distance between data points of both classes)

  • Maximizes margin to create wider gap between the different classes so it’ll work better on new data

Support vectors: Closest data points to the hyperplane

  • Critical role in deciding the hyperplane and margin

If data is not linearly separable, uses cool math to introduce new dimensions that slice the data (kernel)

Has multiple hyperparameters:

  • alpha
  • C: Regularization parameter, inversely proportional to regularization strength
  • Kernel: Mathematical method to transform data into higher-dimensional space
    • Linear kernel SVMs do well with high dimensional data