A typical ML model
Learns to predict an output, but its internal reasoning can be difficult to inspect or communicate.
Local-first symbolic regression
Turn data into equations.
Equarith discovers compact, inspectable formulas from numerical data—on your machine, ready to validate and export.
No account required · Free for up to 200 rows and 4 input variables
Braking distance · 120 rowsError–complexity frontier
Search running locally
Simple
Accurate
Selected equation
d = 0.28v + 0.005v²
Test RMSE1.842
Complexity9
01
0.28v + 0.005v²1.84202
0.0062v²2.11603
0.46v3.908
CSV / TSVParetoResidualsCheckpoints14 export targets7 SDKs · coming very soon
Symbolic regression, simply
What if your data could explain itself?
Symbolic regression searches for a mathematical rule that connects your inputs to the result you want to understand.
-
01
Start with examples.
Give it columns of measurements: speed, temperature, time, cost—or any numerical variables from your problem.
-
02
It explores equations.
The search combines variables and mathematical operations, tests many formulas and keeps the useful ones.
-
03
Choose the formula that fits.
Instead of one answer, you see accurate and simple candidates, then choose the best balance for your goal.
Prediction or understanding?
Symbolic regression
Returns an equation you can read, test, discuss and reuse without the original search engine.
Both can predict well. Symbolic regression shines when you also need to understand, explain, verify or deploy the relationship.
Equarith Free
Bring your next dataset.
No account required · Free for up to 200 rows and 4 input variables
01
Equations you can inspect.
Every candidate is explicit. Compare formula complexity, training error and test performance before you choose.
Local
Your data stays with you.
Import CSV or TSV data and run the entire search locally. No dataset upload. No remote compute backend.
Pareto
See the trade-off.
Explore a Pareto front instead of accepting one opaque answer. Choose the balance of accuracy and simplicity that fits the job.
01 → 14
From discovery to working code.
Export your selected solution to Python, C, C++, Java, JavaScript, TypeScript, C#, R, MATLAB/Octave, Julia, Excel, Rust or Go.
PythonCC++JavaJavaScriptTypeScriptC#
RMATLABJuliaExcelRustGoOctave
Made for real work
The workbench, classroom and lab.
01
Engineering
Turn test-bench, sensor, process and simulation data into readable models.
↗ 02Makers
Calibrate sensors and turn experiment logs into formulas that run anywhere.
↗ 03Education
Teach fit, complexity, overfitting and validation with visible results.
↗ 04Research
Run reproducible searches and compare nondominated candidates.
↗Equarith Free
Bring your next dataset.
Download Equarith free. No account required.