Errors and residuals

Ever wonder how accurate predictions are measured?

Imagine you're tracking your daily step count with a smartwatch. You know your actual steps, but the smartwatch estimates them. How can you tell if it's giving you a good estimate?

Think of RMSE like a scorecard for how close your smartwatch's predictions are to your real step counts. It's a way to see how often the watch is off by a certain amount.

Example

You took 10,000 steps, but your watch predicted 9,800. If this happens every day for a week, RMSE helps you see how much you're consistently off.

Remember this

RMSE tells you the average size of the smartwatch's errors in predicting your steps.

Text adapted from Wikipedia, licensed under CC BY-SA 4.0.

Related concepts

Mean squared error

Mean squared error (MSE) formula: MSE = (1/n) * Σ(y_i - ŷ_i)²

Standard deviation

Standard deviation (σ) is the square root of variance

RMSprop fixes about AdaGrad: uses exponential moving average instead of sum

RMSprop uses an exponentially decaying average of squared gradients, unlike AdaGrad's cumulative sum

Brier score

Brier score measures mean squared error of probability predictions

Regression analysis

Linear regression equation: ŷ = β0 + β1X

calibration means: a model predicting 80% should be correct 80% of the time

How can we trust a coin flip to predict the weather?

Swipe through 100 ML concepts daily

Open Pocket Polymath