Welcome to Neurogebra¶
The Executable Mathematical Formula Companion for AI and Data Science¶
v2.5.8 | Curated Formula Library | Base + Expanded Dataset Loaders | Training Observatory Pro
Neurogebra is a unified Python library that bridges symbolic mathematics, numerical computation, and machine learning. It provides a curated expression library, base and expanded dataset loaders, and a training system with full mathematical transparency -- designed for students, researchers, and engineers alike.
Who is this for?
- Students learning ML/AI who want to see the math behind every operation
- Researchers who need reproducibility, rapid formula prototyping, and transparent diagnostics
- Engineers who want a verified formula library with production-ready logging
What Makes Neurogebra Different?¶
| Feature | Traditional Frameworks (PyTorch, TF) | Neurogebra |
|---|---|---|
| Learning curve | Steep -- many hidden abstractions | Gentle -- every step is explained |
| Math visibility | Hidden inside C++ kernels | Symbolic -- you SEE the formulas |
| Expressions | Tensors/Modules you don't understand | Mathematical expressions you can read |
| Explanations | Read research papers | Built-in .explain() on everything |
| Training diagnostics | Basic loss curves | Full math transparency with Observatory Pro |
| Reproducibility | Manual tracking | Automatic training fingerprinting |
| Formula library | Build your own | Curated, searchable, composable formulas |
| Target audience | Production engineers | Students, researchers, and engineers |
Quick Example¶
from neurogebra import MathForge
# Create the main interface
forge = MathForge()
# Get the ReLU activation function
relu = forge.get("relu")
# See what it actually IS
print(relu.explain())
# Output: "ReLU (Rectified Linear Unit) outputs x if x > 0, else 0"
# See the formula
print(relu.formula)
# Output: Max(0, x)
# Evaluate it
print(relu.eval(x=5)) # 5
print(relu.eval(x=-3)) # 0
# Get its derivative (gradient)
relu_grad = relu.gradient("x")
print(relu_grad)
See how readable that is?
Every expression tells you what it is, how it works, and why it matters. No magic. No black boxes.
What You'll Learn in This Documentation¶
This documentation is structured as a progressive learning path, starting from absolute basics:
Getting Started¶
Install Neurogebra and write your first program in under 5 minutes.
Python for ML (Refresher)¶
A quick refresher on Python basics, NumPy, and data handling -- the prerequisites for ML.
ML Fundamentals¶
What Machine Learning actually is, the types of ML, the standard workflow, and the math behind it all.
Neurogebra Tutorial¶
Step-by-step lessons on every feature -- from expressions to training to model building.
Training Observatory and Observatory Pro¶
Real-time training diagnostics with adaptive logging, automated health warnings, epoch summaries, visual dashboards, and training fingerprinting.
Advanced Topics¶
Custom expressions, framework bridges (PyTorch / TF / JAX), visualization, and optimization.
Full Projects (Neurogebra vs PyTorch)¶
3 complete ML/Deep Learning projects with side-by-side code comparison between Neurogebra and PyTorch.
Real Project Notebooks (Colab)¶
10 intermediate to advanced, real-dataset notebooks for image classification, GANs, diffusion workflows, core NLP, and small language model training. Each notebook includes official dataset source links and explicit download instructions.
Install Now¶
Ready? Start with Installation →