Q1 model representation
Q2: cost function
For regression problems common cost function is squared error cost function:
We aim to select the appropriate parameters Θ such that the error function is minimized, i.e., a straight line most approximate the real situation.
Q3: an intuitive understanding of the cost function (a parameter)
Q4: intuitive understanding (two parameters) of the cost function
Q5 gradient descent
Q6 gradient descent intuitive understanding
(1) the gradient descent method of any cost function can be minimized, but not limited to linear regression cost function.
(2) local minimum when getting close, a gradient value becomes small even if the learning rate constant, changes in amplitude parameter also decreases.
(3) the learning rate is too small parameter changes slowly, to reach the optimum length of time, when large learning rate, the cost function may lead to non-convergence, or even divergence.
(4) a gradient of a point is the slope.
Linear regression gradient descent Q7
English vocabulary
Linear regression with one variable --- univariate linear regression model representation --- represents a model training set --- training set hypothesis assumed --- gradient descent --- gradient descent Convergence --- convergence local minimum --- local minimum value --- global maximum global minimum
Univariate linear regression Course summary (point I)