Dictionary Based Pattern Entropy for Causal Direction Discovery

Collection of arXiv stat_ml preprints (2026-03-06) covering new methods and theoretical results across causal discovery, time-series feature maps, probabilistic modeling, and learning theory. Key contributions include: Dictionary Based Pattern Entropy (DPE) for causal direction discovery combining algorithmic and Shannon information; Bayesian hierarchical and mechanistic models for Collatz stopping times; the Volterra signature as an explicit, kernelized feature map for history-dependent systems; prediction-augmented optimal testers for distributional independence; analyses of convolutional NN