Machine learning (--regularization:regularized) Linear regression

Machine learning (--regularization:regularized) Linear regression

• Gradient descent
  • Without regularization

• • With regularization

• • • Θ0 is the same as the original, no regularization.
    • The θ1-n is slightly smaller than the original (1-αλ⁄m) <1

• Normal equation
  • Without regularization

• • With regularization

• In normal equation, when XTX is not reversible

• • If m<=n, there will be a case of matrix irreversible
  • Although PINV can be calculated when X is not reversible in octave, if the language used or the inv is used, it cannot be calculated when X is not reversible.
  • Regularization considering this, when +, the matrix must be reversible.

Machine learning (--regularization:regularized) Linear regression