Symbolic Regression
Fit compact, interpretable equations to a CSV. Everything runs in your browser.
Run
Data
Example datasets
- Semi-major axis a (AU), sidereal orbital period T (years) and eccentricity e of 21 bodies orbiting the Sun: the eight planets, dwarf planets, large asteroids and comets. Expect T ≈ a^1.5; e can be ticked as a distractor.
- 300 synthetic pairs of bodies: masses m1, m2 in 10²⁴ kg, distance r in 10⁹ m, force F in 10¹⁸ N, with 1.5% noise. Expect F ≈ 66.74·m1·m2/r²; temperature_K is a distractor.
- 300 synthetic parcels: weight in kg, delivery zone and carrier, price with 2% noise. Expect cost ≈ 4 + 0.8·weight + 6·1[zone ∈ {island, remote}]; carrier is a distractor.
Settings
Share of rows, picked at random by the seed, held back to score the equations. Constants are never fitted to them.
Results
Load data, pick a target and press Run. The equations found appear here, from the simplest to the most accurate.
How it works
Load a CSV or pick an example, choose the column to predict and the inputs, then press Run. The search keeps the most accurate equation at each size. The one marked best balances accuracy against size. Select an equation or a point on the plot to inspect it.
Your data stays in your browser. The details are in the docs.