- A method to find the minimum or maximum of a function by moving along the 

    slope by a positive or negative direction according to the sign of the 

    derivative

- Process / workflow:

    - goal: keep descending until the loss is minimized. Stop the process 

        when the loss start increasing.

    - randomly take a number and calculate slope parameter

    - move the slope to a positive direction by adding a small number

    - move the slope to a negative direction by subtracting a small number

    - How large the number should be added is determined by a parameter called l

        earning rate

        - too small learning rate = hardly any improvement in each iteration

        - too large learning rate = sometimes we will miss the minimum

    - calculate equation after each move to see which iteration provides the least loss 

        (the minimum loss, global extrema)