About
A research organization built on academic rigor and high-performance compute.
Valence Quantitative Research LP conducts applied research in statistical machine learning. Our group works on estimation and inference problems that arise when data is high-dimensional, non-stationary and noisy. We publish technical notes, engineering write-ups and methodological tools in the open, and we measure our own work against the standards of peer-reviewed statistics.
- Founded
- 2007
- Location
- Berkeley, California
- Focus
- Statistical ML, state-space models, high-dimensional inference
Principles
Derivation before heuristics
Every method we publish is stated with its assumptions, its derivation, and the conditions under which it fails.
Rigor in low signal-to-noise settings
Many of the hardest problems in statistics have signals that are barely above the noise floor. We design estimators for that regime explicitly.
Reproducibility as a requirement
Code, configurations and seeds accompany our technical notes so that results can be independently re-derived.
What we publish
Our publications are educational and technical. They describe methods, mathematical results and software. They are not investment advice, and they are not offered as a basis for any financial decision. Each article includes a BibTeX citation generator so that readers can reference our work accurately.