Founded 2025 · Open source · Python
Machine Gnostics
The world's first machine learning library built on non-statistical principles — encoding the laws of nature directly into algorithms.
The problem I set out to solve
Working at the intersection of thermal engineering and AI, I kept running into the same wall. Real industrial and reserach data is small, noisy, and messy. Sensors fail. Experiments are expensive. You rarely get thousands of samples. Yet every mainstream AI framework — from scikit-learn to PyTorch — is built on statistical assumptions that silently break down when data is scarce or corrupted.
The deeper issue was not just sample size. It was that statistical AI has no concept of physical reality. It treats data as abstract numbers drawn from probability distributions, with no grounding in geometry, thermodynamics, or the measurable laws that govern the real world.
"Why are we forcing physical data into statistical models that were never designed for it?"
That question, which I first encountered during my PhD research at the Czech Academy of Sciences in Prague, became the seed of Machine Gnostics.
Standing on the shoulders of a pioneer
Machine Gnostics did not emerge from nothing. It is built on the life's work of Dr. Pavel Kovanic (1942–2023), a researcher at the Institute of Information Theory and Automation of the Czechoslovak Academy of Sciences, who first published Mathematical Gnostics in 1984.
Kovanic's radical idea: uncertainty in data is not random — it has measurable, material causes. Instead of statistical distributions, he encoded Riemannian geometry, thermodynamic entropy, and relativistic mechanics into a framework that lets data speak for themselves.
His theory was met with scepticism for decades. In 2022, during my PhD at UCT Prague under Dr. Magdalena Bendová and Dr. Zdeněk Wagner — both long-time collaborators of Kovanic — I identified the bridge between Mathematical Gnostics and modern machine learning. In 2023, the year Kovanic published his final book and passed away, Machine Gnostics was born.
It is my honour to carry his work forward.
What Machine Gnostics actually does
Unlike statistical AI, Machine Gnostics encodes the laws of nature — geometry, physics, entropy — directly into its algorithms. This produces three properties that statistical frameworks cannot replicate:
Works on Small Data
No minimum sample size. No assumption of normal distributions. Gnostic algorithms extract reliable signal even from 10–20 data points — the kind of datasets that are routine in engineering, medicine, and industrial process control.
New Explainability
Every result traces back to geometry, physics, and thermodynamics without statistical assumptions — fully interpretable and auditable for regulated industries and scientific research.
Noise and Outlier Robust
Because the framework treats each data point as a material event — not a statistical sample — it naturally distinguishes signal from noise without discarding outliers or requiring data cleaning assumptions.
What's inside the library
Gnostic - Data Analysis
Exploratory analysis, distribution functions (EGDF, ELDF, QGDF, QLDF), clustering, interval analysis, homogeneity tests — all without statistical assumptions.
Gnostic - Machine Learning
Classification, regression, clustering, and time series forecasting models grounded in gnostic certainty. Fully traceable and explainable.
MAGNET - Deep Learning
Next-generation neural networks built on mathematical gnostic principles. Noise-immune. Thermodynamically grounded. Easy Pythonic API.
Get started in one line
pip install machinegnostics
Full documentation, tutorials, and Jupyter notebook examples are available at docs.machinegnostics.com.
Collaborate with me
I actively welcome research collaborations that apply Machine Gnostics to new domains — particularly in engineering, medicine, environmental science, and industrial AI. If you are working on a problem where conventional statistical AI is failing, this framework may be exactly what you need.
Research Partnerships Joint research, co-authorship, and applying gnostic methods to novel scientific problems.
Open Source Contributions Contribute new algorithms, modules, documentation, or domain-specific implementations.
Master, PhD & PostDoc Supervision Supervising graduate researchers who want to work at the intersection of Mathematical Gnostics and AI.
Industry Application Applying Machine Gnostics to real industrial data challenges where small-sample or noisy data is the norm.
Frequently Asked Questions
What makes Machine Gnostics different from statistical machine learning?
Traditional ML (scikit-learn, PyTorch) relies on statistical assumptions: your data must come from probability distributions, you need large sample sizes (typically thousands), and uncertainty is treated as random noise.
Machine Gnostics encodes geometry, physics, and thermodynamics directly. It treats uncertainty as material cause, not randomness. This means:
- Works with 10-20 data points instead of thousands
- No statistical assumptions needed
- Every result is fully traceable to physical laws
- Results are interpretable for regulated industries
Can I use Machine Gnostics for my problem?
Machine Gnostics excels when:
- Your dataset is small (< 100 samples typical)
- Data is noisy or corrupted (sensors fail, experiments are expensive)
- You need new explainability (regulated industries, scientific research)
Connect with me if you're unsure.
How do I get started?
Install with one line:
pip install machinegnostics
Then explore:
- Quick start guide: docs.machinegnostics.com
- Jupyter tutorials: Available in the docs
- GitHub examples: github.com/MachineGnostics
Is Machine Gnostics open source?
Yes! Machine Gnostics is fully open source. Contributions welcome on GitHub.
Can I collaborate or contribute?
Absolutely. I welcome:
- Research partnerships on new applications
- Open source contributions (algorithms, documentation, examples)
- PhD/postdoc supervision at the intersection of gnostic ML and your domain
Get in touch with your ideas!
Acknowledgements
Machine Gnostics stands on the foundation laid by Dr. Pavel Kovanic (1942–2023), whose vision of a nature-grounded, non-statistical theory of data uncertainty inspired this entire project. His legacy lives on in every algorithm.
Deep thanks also to Dr. Magdalena Bendová (PhD supervisor) and Dr. Zdeněk Wagner (expert supervisor) at the Czech Academy of Sciences, Prague — whose mentorship, encouragement, and decades of work with Mathematical Gnostics made this journey possible.
Full history: click here!