Quickstart

Go from formula to trained model in 60 seconds

💡Tip

MathExec gets you from formula to trained model in under 60 seconds. This guide walks through the three core steps.

Step 1: Write a Formula

Open the app and select the Formula tool (press F). Click anywhere on the canvas and type a LaTeX formula:

y = \sigma(Wx + b)

This creates a formula block. MathExec recognizes it and compiles it to a trainable PyTorch model.

ℹī¸Note

You can also draw formulas by hand using the Draw tool (P). MathExec uses OCR to recognize your handwriting.

Step 2: Add Data

Switch to the Data tool (D) and click on the canvas. Pick a dataset from the popup — you can upload a CSV, import from Kaggle, or use a sample dataset.

A green data block appears on the canvas. Connect it to your formula block by drawing a line between their ports.

Step 3: Train

Press Cmd+Enter (or Ctrl+Enter on Windows) to start training. The training drawer slides open showing:

  • Live loss curve updating each epoch
  • Metrics — accuracy, F1, precision, recall (classification) or MSE, R2 (regression)
  • Export button — download the generated PyTorch code

That's it. Three steps, under a minute.

What's Next?

  • Canvas — learn all the tools and block types
  • Training — configure hyperparameters, optimizers, and loss functions
  • Formulas — full reference of supported formulas and architectures
  • Data Studio — clean and transform your data before training
NextCanvas