Statistical models with time-varying parameters

Quem: 

Henrique Helfer Hoeltgebaum

Onde: 

Praia de Botafogo, 190 - sala 537

Quando: 

6 de Maio de 2019 às 16h

This work is composed of three papers in which the common ground among them is statistical models with time-varying parameters. All of them adopt a framework that uses a data-driven mechanism to update its coefficients. The first paper explores the application of a new class of non-Gaussian time series framework named Generalized Autoregressive Scores (GAS) models. In this class of models the parameters are updated using the score of the predictive density. We motivate the use of GAS models by simulating joint scenarios of wind power generation. Considering three wind power plants located in the Northeast of Brazil. As argued, such scenarios are highly demanded for the development of many power-system planning, operation and generation investment under the uncertainty of renewable energy generation.

In the last two papers, Stochastic Gradient Descent (SGD) is adopted to update time-varying parameters. This methodology uses the derivative of a user specified cost function to drive the optimization. The developed framework is designed to be applied in a streaming data context, therefore adaptive filtering techniques are explored to account for concept-drift. We explore this framework on cyber-security and instrumented infrastructure applications.

*Texto informado pelo autor. 

Palestrante: 

Henrique Helfer Hoeltgebaum graduated from the Federal University of Rio Grande do Sul (UFRGS), Brazil, with a B.S. in Statistics. He was fully funded during both MSc and PhD at the Electrical Engineering Department of PUC-Rio, Brazil, with emphasis in time series and optimization. Currently he received another scholarship from the Brazilian government, due to outstanding research contributions, to work one year applying machine learning models to big data streams under the supervision of Professor Niall Adams at Imperial College London. For the past three years he has been working with streaming data, developing anomaly detectors to make decisions in real-time as data arrives sequentially. Currently he holds a position as Research Assistant at the Alan Turing Institute, working in the intersection between machine learning methods and instrumented infrastructure, specifically applied to nuclear energy.