Data Visualization
Curve Fitter
An interactive polynomial fitting sketch with generated data, pan/zoom graphing, finite-difference coefficient updates, and visible error feedback.
Employer signal
What This Project Shows
This project turns optimization into something visible. It shows how a model changes when coefficients move, rather than hiding the process behind a library call.
Problem
What Needed To Be Solved
Regression can feel like a black box when the only output is a final curve. A useful teaching or debugging tool needs to expose how coefficients, error, and data interact.
Approach
How I Built The Solution
I built the fitting loop manually. The sketch computes mean error, perturbs coefficients to estimate local improvement direction, and updates the polynomial while rendering both data and model in a pannable coordinate system.
Outcome
What It Demonstrates
The project demonstrates numerical reasoning, interface feedback, and comfort explaining an algorithm through the UI itself.
Evidence From Source
Source signal
The p5 source uses `error_function(data, coefficients)`, perturbs individual coefficients, and re-renders the polynomial with `show_polynomial(...)`.
Design value
Writing the fit loop manually made the tradeoff visible: better for learning and debugging than a one-line regression API.