Course URL: Machine Learning | Coursera
The main content of the second week has two parts: multiple linear regression and Octave tutorial.
1. Multiple Linear Regression
1. Basic model
2. Multivariate Gradient Descent
3. Feature Scaling
Control the range of each feature approximately in [-1,1]
Mean normalization:
4. Learning rate
If it is too small, gradient descent will be slow.
If it is too large, the cost function may not decrease with each iteration and may not converge.
Tried with triple increments, ..., 0.001, 0.003, 0.01, 0.03, 0.1, 0.3, 1, .....
5. Normal equation solution method
Octave：pinv(X' * X)*X' * y
6. Comparison of gradient descent and normal equation method
When n is too large, it is more inclined to use the gradient descent algorithm.
Two, Octave tutorial (omitted)