kernels

Kernels I

有沒有更好的方式來算　features f1, f2, f3 …

Kernels II

大神　takeaway: 不要自己刻 SVM!!　請拿好心叔叔伯伯姐姐阿姨提供的 package 來使用~

• Chossing the landmarks

• SVM parameters:

  • C (= 1 / λ).
    • Large C: Lower bias, high variance.
    • Small C: Higher bias, low variance.
  • σ^2

    • Large σ^2: Features fi vary more smoothly.
      Hiner bias, lower variance.

    • Small σ^2: Features fi vary less smoothly.
      Lower bias, highter variance.

  

  • EX:
    Suppose you train an SVM and find it overfits your training data. Which of these would be a reasonable next step?

    • Decrease C
    • Increase σ^2