Information theory · Quantitative research

Anatoly Vitold Stankyavichyus

Senior Data Scientist — varentropy, time-series modeling, and tail-risk measurement.

Varentropy

\[ V(X)\;=\;\operatorname{Var}\!\big(-\log f(X)\big) \]

Student t ν = 2

scale σ = 1

density f(x)

density of surprisal −log f(X)

H(X) = — V(X) = — √V = —

\[ V(X)=\Big(\tfrac{\nu+1}{2}\Big)^{2}\big[\,\psi'(\tfrac{\nu}{2})-\psi'(\tfrac{\nu+1}{2})\,\big] \]

ν: V ∼ 1/ν² → ∞ as ν → 0⁺ · = π²/3 at ν = 1 (Cauchy)

σ: H(σX) = H(X) + log σ (entropy shifts) · V(σX) = V(X) (scale‑invariant)

About

Anatoly Vitold Stankyavichyus is a Senior Data Scientist working in the utilities sector specializing in load research, time-series modeling, and large-scale energy data analytics. At PSEG Long Island, he has developed and productionized models for load disaggregation, weather normalization, demand response performance assessment, rate design and customer load profiling at scale. His work combines advanced statistical methods with modern data engineering to deliver practical insights across millions of customers.

M.S. in Applied Mathematics and Statistics, Stony Brook University.

Preprints

Varentropy: Overview, Computational Routes, and Structural Decomposition

Overview of varentropy and a useful decomposition for constructing lower bounds.

PDF LaTeX source

On Varentropy of Stable Laws with Rational Stability Index

D-algebraic desciption of densities of α-Stable laws with rational α, formula for varentropy, and Varentropy CLT.

PDF LaTeX source Code

The Price of Jumpiness: Varentropy, Magnitude-Information Proﬁles, and Finite-Horizon Floor-Breach Risk in Kelly Allocation

Application of Varentropy to Kelly Allocation.

PDF LaTeX source Code

Elsewhere

ORCID LinkedIn