#0011

One Step of Gradient Descent

MediumML20 min15 XP

Problem

Given a current weight, a gradient, and a learning rate, return the updated weight after one gradient descent step.

Why This Matters

Many ML algorithms are just repeated parameter updates. This problem makes the update rule concrete.

Function Signature

def gradient_step(weight, gradient, learning_rate):

Examples

Example 1

Inputweight = 2.0, gradient = 0.5, learning_rate = 0.1

Output1.95

Move opposite the gradient: 2.0 - 0.1*0.5 = 1.95.

Constraints

Inputs are numbers.
Learning rate is non-negative.
Return a number.

CodePython

def gradient_step(weight, gradient, learning_rate):
    """
    Input example: weight = 2.0, gradient = 0.5, learning_rate = 0.1
    Return: 1.95
    Return the value from the function so the tests can compare it. Do not print unless the prompt asks.
    """
    # write your code here
    return weight

def gradient_step(weight, gradient, learning_rate):
    """
    Input example: weight = 2.0, gradient = 0.5, learning_rate = 0.1
    Return: 1.95
    Return the value from the function so the tests can compare it. Do not print unless the prompt asks.
    """
    # write your code here
    return weight

Visible browser tests run here when available.

Testcases2 visible / 4 hidden categories

Positive gradient

Input2.0, 0.5, 0.1

Expected1.95

A positive gradient decreases the weight.

Negative gradient

Input2.0, -0.5, 0.1

Expected2.05

A negative gradient increases the weight because subtracting a negative adds.

Hidden Test Categories

Zero gradientZero learning rateLarge gradientFloating point tolerance