Dennis Alexis Valin DittrichMar 9

Robust estimation demands highly efficient computation, especially in streaming anomaly detection where latency budgets are tight.

While Rousseeuw & Croux's robust estimators ($Q_n$ and $S_n$), and Rousseeuw & Verboven's M-estimators of location and scale for very small samples, provide exceptional reliability, computing them requires intensive math.

robscale 0.1.5 is now on CRAN. It delivers a native C++17/Rcpp implementation designed for absolute speed. The package utilizes SIMD-vectorized $\tanh$ evaluation, Newton-Raphson iteration, and optimal sorting networks for cache-aware median selection.

The result? A 1.6x up to ~28x performance leap over pure-R implementations. The mathematical results remain identical; only the computational underpinnings change.

📦 CRAN: https://cran.r-project.org/package=robscale
💻 Code: https://github.com/davdittrich/robscale

#RStats #RobustStatistics #DataScience #Optimization

Claudio AgostinelliJan 25
Our paper on "Hellinger loss function for Generative Adversarial Networks" is posted in arXiv at http://arxiv.org/abs/2512.12267
#Statistics
#HellingerDistance
#NeuralNetworks
#GenerativeAdversarialNetworks
#RobustStatistics
#InfluenceFunction
Hellinger loss function for Generative Adversarial Networks

We propose Hellinger-type loss functions for training Generative Adversarial Networks (GANs), motivated by the boundedness, symmetry, and robustness properties of the Hellinger distance. We define an adversarial objective based on this divergence and study its statistical properties within a general parametric framework. We establish the existence, uniqueness, consistency, and joint asymptotic normality of the estimators obtained from the adversarial training procedure. In particular, we analyze the joint estimation of both generator and discriminator parameters, offering a comprehensive asymptotic characterization of the resulting estimators. We introduce two implementations of the Hellinger-type loss and we evaluate their empirical behavior in comparison with the classic (Maximum Likelihood-type) GAN loss. Through a controlled simulation study, we demonstrate that both proposed losses yield improved estimation accuracy and robustness under increasing levels of data contamination.

arXiv.org
Claudio AgostinelliJul 8, 2025
Our paper on "Robust penalized estimators for high-dimensional
generalized linear models" is accepted in TEST, see it at https://link.springer.com/article/10.1007/s11749-025-00978-6
or in arXiv at https://arxiv.org/abs/2312.04661
#HighDimension
#GeneralizedLinearModels
#PenalizedMethods
#RobustStatistics
#MT-Estimators
Robust penalized estimators for high-dimensional generalized linear models - TEST

Robust estimators for generalized linear models (GLMs) are not easy to develop due to the nature of the distributions involved. Recently, there has been growing interest in robust estimation methods, particularly in contexts involving a potentially large number of explanatory variables. Transformed M-estimators (MT-estimators) provide a natural extension of M-estimation techniques to the GLM framework, offering robust methodologies. We propose a penalized variant of MT-estimators to address high-dimensional data scenarios. Under suitable assumptions, we demonstrate the consistency and asymptotic normality of this novel class of estimators. Our theoretical development focuses on redescending $$\rho $$ ρ -functions and penalization functions that satisfy specific regularity conditions. We present an Iterative re-weighted least-squares algorithm, together with a deterministic initialization procedure, which is crucial since the estimating equations may have multiple solutions. We evaluate the finite-sample performance of this method for Poisson distribution and well-known penalization functions through Monte Carlo simulations that consider various types of contamination, as well as an empirical application using a real dataset.

SpringerLink
Claudio AgostinelliFeb 19, 2025
Registration is open for the summer school "Robust Statistics: Theory and Computation" to be held in Ispra (Varese), on 15-17 May 2025.
Information at
https://datascience.maths.unitn.it/icors2025/school.html
#RobustStatistics
#SummerSchool
Summer School

daTa scieNce is the web site of the students in Mathematics for daTa scieNce at the Departement of Mathematics, University of Trento

Claudio AgostinelliJan 24, 2025
Registration and submission of abstract are open for the International Conference on Robust Statistics 2025 (ICORS2025).
All the information at
https://datascience.maths.unitn.it/icors2025
#RobustStatistics
#ICORS2025
International Conference on Robust Statistics 2025

daTa scieNce is the web site of the students in Mathematics for daTa scieNce at the Departement of Mathematics, University of Trento

Claudio AgostinelliDec 26, 2024
14 February 2025 seminar by Peter Filzmoser on Outlier identification and explanation for matrix-valued observations. Full information at https://datascience.maths.unitn.it/events/oi2025/index.html
#RobustStatistics
#daTascieNce
Outlier identification and explanation for matrix-valued observations

daTa scieNce is the web site of the students in Mathematics for daTa scieNce at the Departement of Mathematics, University of Trento