Engineering Sandbox rollout · bundle-mounted ML essays

Which line makes the squared errors smallest?

Change the line's coefficients and compare the residuals with the loss readout. Squaring gives large errors more weight; a line that looks close to most points may still have a large loss. A good fit here does not establish causation or test accuracy.

Probe a prediction Tune the loss directly Watch gradient descent converge

How to use this page

In the loss scene, change one coefficient while holding the other fixed and look for a lower loss. Then follow gradient descent: its updates use the loss slope rather than a visual guess. Compare the optimizer's path with your manual adjustments.

Play first: fit the line Jump to optimization

Loading Linear Regression lesson…