Algorithms for quantifying associations, independence testing and causal inference from data.
-
Updated
Jul 26, 2026 - Julia
Algorithms for quantifying associations, independence testing and causal inference from data.
A 30-second test of whether a pattern-finder can tell a real signal from random noise, most can't. An open finding from my ML research on financial markets.
Non-stationarity breaks permutation surrogates in multi-agent RL: analysis code and data
Code, data and paper: a polymer whose stiffness follows prime-family gaps reads only sieve composition and the density envelope, and is blind to congruence structure. Exact ideal-chain theory, Granville sieve models, Langevin tests.
Методология Геометрического Анализа Рынка. Классификация рыночных состояний через фазовое пространство.
Code, data and paper: the prime-gap imprint on a self-avoiding polymer survives full relaxation, but its size-proportional correlation length is a property of the readout, reproduced by surrogates without arithmetic. Langevin and exact ideal-chain analysis.
Benchmark testing whether catch22 time-series features can detect genuine chaos even when linear statistics (mean, variance, spectrum) are deliberately matched via IAAFT surrogate decoys.
To associate your repository with the surrogate-data topic, visit your repo's landing page and select "manage topics."