Traditional comparative methods struggle when attestations are fragmentary. We place sound-change priors in a hierarchical model fit with Hamiltonian Monte Carlo on small cognate tables from two Japonic-related lects. Posterior consonant inventories align with expert proposals while quantifying uncertainty on vowel mergers that manual reconstructions left implicit.

Bivalves integrate particle loads over weeks yet laboratory exposure protocols rarely match river turbulence. We quantify fibers and fragments in unionids along an agricultural gradient and relate tissue burdens to clearance rates measured in field mesocosms. Elevated counts correlate with reduced feeding efficiency independent of dissolved oxygen, suggesting sublethal stress before mortality events.

High-albedo surfaces can lower radiant loads but durability and skid resistance limit adoption. We monitor coated and control asphalt strips for two summers with embedded thermistors and pedestrian thermal comfort surveys. Coatings that retain seventy percent of initial albedo after one season reduce afternoon surface temperatures measurably without increasing wet-night friction risk beyond local standards.

Spacing between heaving buoys alters constructive and destructive interference that simple linear models understate. We couple a boundary-element hydrodynamic solver with a genetic algorithm over heading and separation grids using measured spectra from two Pacific test sites. Optimized staggered layouts raise estimated annual capture by twelve to eighteen percent versus inline defaults at similar mooring cost.

Sampling pigments damages fragile folios, yet single-instrument scans can misclassify layered inks. We register co-located XRF and Raman maps with a lightweight fusion network trained on micro-sampled reference tiles. Blind tests on fourteenth-century choir books recover azurite versus indigo zones with fewer false positives than peak-matching rules alone.

Inhaled corticosteroid adherence remains low despite standard education. We cluster-randomize families to static versus adaptive reminders that adjust timing from prior missed-dose patterns. Over sixteen weeks, adaptive arms show higher controller use and fewer rescue inhaler events, with largest gains among households reporting irregular sleep schedules at baseline.

Classical mean-variance rebalancing ignores cumulative fees when horizons stretch across quarters. We encode cardinality and turnover penalties in a quadratic unconstrained binary optimization form solvable by hybrid quantum-inspired annealers and compare against greedy baselines. On historical equity panels, the heuristic reduces turnover while keeping tracking error within pre-set bands relative to a buy-and-hold benchmark.