Insights & Tutorials
The MathExec Blog
Deep dives into formula compilation, model training, and the future of mathematical computing.
Formula Complexity vs. Model Performance: Do More Complex Equations Actually Train Better?
We trained 6 progressively complex formulas on the same dataset. The results surprised us.
The Formula Compiler Problem: Why Parsing Math is Harder Than Parsing Code
Programming languages have grammars. Mathematical notation has conventions, context, and centuries of overloaded symbols. Building a formula compiler means solving all of that.
We Compiled 17 Textbook Formulas to PyTorch. Here's What Broke.
We ran 17 standard ML formulas through MathExec's compiler. 10 compiled cleanly. 7 didn't. The failures taught us more than the successes.
LaTeX to PyTorch: A Reference Taxonomy of Formula Patterns and What They Compile To
A definitive reference mapping LaTeX formula patterns to their PyTorch equivalents. Every entry backed by MathExec's compiler.
The Math-to-Code Translation Tax: What Gets Lost When Formulas Become Software
The formula in the paper is never the whole story. Batch dimensions, numerical stability tricks, weight initialization, shape inference: the hidden work that textbooks don't mention.
We Counted: The Average ML Paper Formula Takes 94 Lines of PyTorch
We took 10 formulas from ML papers and textbooks and wrote the complete PyTorch equivalent for each. The average was 94 lines. The formula was never more than 40 characters.
60-Second Model: Benchmarking Time-to-First-Prediction Across 5 ML Workflows
We timed 5 different ML workflows on the same task: raw PyTorch, Lightning, scikit-learn, AutoML, and MathExec. The gap was bigger than we expected.
Why We Chose Formulas Over Drag-and-Drop for ML
Most ML platforms use drag-and-drop node editors. We chose mathematical notation instead.
Building a Data Studio: How We Added NL-Powered Data Transforms
How we built MathExec's Data Studio, a feature that lets users transform datasets using plain English instructions powered by LLMs.
5 Machine Learning Models You Can Train with One Line of Math
From linear regression to neural networks: 5 models you can train in MathExec just by typing a formula.
How MathExec Compiles LaTeX to PyTorch
A deep dive into MathExec's formula compiler: how we parse LaTeX expressions and generate equivalent PyTorch modules with trainable parameters.
Introducing MathExec: From Formula to Trained Model in 60 Seconds
MathExec lets you write a math formula, compile it to PyTorch, train on your CSV data, and export production-ready Python code. All in under a minute.