At each step, we know how to calculate the gradient and move in that direction with a step size proportional to our learning rate. Now, we want to make these steps until we reach convergence.
Instructions
We have all of the functions we have defined throughout the lesson.
Now, let’s create a function called gradient_descent() that takes in x, y, learning_rate, and a num_iterations.
For now, return [-1,-1].
In the function gradient_descent(), create variables b and m and set them both to zero for our initial guess.
Return b and m from the function.
Update your step_gradient() function to take in the parameter learning_rate (as the last parameter) and replace the 0.01s in the calculations of b_gradient and m_gradient with learning_rate.
Let’s go back and finish the gradient_descent() function.
Create a loop that runs num_iterations times. At each step, it should:
- Call
step_gradient()withb,m,x,y, andlearning_rate - Update the values of
bandmwith the valuesstep_gradient()returns.
Outside of the function, uncomment the line that calls gradient_descent on months and revenue, with a learning rate of 0.01 and 1000 iterations.
It stores the results in variables called b and m.
Uncomment the lines that will plot the result to the browser.