We study problems where observations are noisy, incomplete, high-dimensional, or non-stationary. Current published work applies these methods to tropical-Pacific climate forecasting and financial time series.

Research practice

Our work emphasizes explicit assumptions, reproducible experiments, meaningful baselines, sensitivity analysis, and careful separation between empirical evidence and interpretation.

Negative and null results are published when they are informative. The research archive includes controlled comparisons in which added model complexity did not improve forecasting, and studies of why an apparent gain failed to transfer.

Research software

Software is developed when an experiment needs new tools, when automation improves reproducibility, or when a method warrants a reusable implementation. Software companions describe the implementations behind the studies.

Collaboration

Deep Wave occasionally collaborates with researchers and organizations on problems aligned with its research programme. Contact