Multiple Features (multi-feature amount)
1, characters are defined:
2, Multivariate Linear Regression (linear regression):
3, the cost function:
4, gradient descent algorithm:
A gradient descent technique --Feature Scaling (feature scaling)
1, define: scaling each feature amount between -1 and 1. (With a range close to, such as 0 to 3, -2 to 0.5, etc.)
2, the role: to facilitate convergence upon a gradient descent (round profile closer).
3, scaling method:
(1) divided by the maximum value directly.
(2) Mean Normalization (mean normalization): subtracting the average divided by the range (i.e., maximum - minimum)
Gradient descent Skills II --Learning Rate (learning rate)
A: α = 0.1 B: α = 0.01 (α is small, convergence speed) C: α = 1 (α is too large, divergent)
Polynomial Regression (polynomial regression)
1, for example: different order terms can be converted to a different feature values, feature values such as the area of x1, characterized ² area value x2, characterized ³ area value x3. The polynomial regression problem into multiple linear regression.