• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Master's Dissertation
DOI
https://doi.org/10.11606/D.45.2021.tde-12052021-114559
Document
Author
Full name
Victor Junji Takara
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
São Paulo, 2021
Supervisor
Committee
Esteves, Luís Gustavo (President)
Diniz, Marcio Alves
Stern, Rafael Bassi
Title in Portuguese
Inferência para o modelo Bernoulli na presença de adversários
Keywords in Portuguese
Aprendizado de máquina
Inferência bayesiana
Teoria da decisão
Abstract in Portuguese
A teoria da decisão com adversários se originou na tentativa de solucionar problemas na área de aprendizado de máquina. Nessa teoria, supõe-se a existência de adversários que têm como intuito a perturbação dos dados (ou do mecanismo gerador dos mesmos). Uma vez que ela é baseada em inferência bayesiana, a todas as incertezas são atreladas medidas de probabilidade, inclusive às possíveis ações realizadas por adversários. No entanto, pela natureza aplicada da teoria, ela foi criada e estudada com enfoque na teoria da decisão, sem muita preocupação com formalismos na área de estatística. Assim, o objetivo desse trabalho foi estudar elementos inferenciais importantes, como a estimação pontual e o teste de hipóteses para o modelo Bernoulli na presença de adversários. Ilustramos como essas alterações impactam a estimativa pontual e o teste de hipótese bayesiano, além da própria distribuição dos dados observáveis e de componentes importantes, como o comportamento do risco bayesiano e regiões críticas.
Title in English
Inference for Bernoulli model in presence of adversaries
Keywords in English
Bayesian inference
Decision theory
Machine learning
Abstract in English
The adversarial decision theory originated in the attempt to solve issues in the area of machine learning. In this theory, it is assumed that there are adversaries whose intention is to disturb the data (or the mechanism which generates them). Since it is based on Bayesian inference, probabilities measures are attached to all uncertain quantities and to possible actions taken by opponents. However, due to the applied nature of this theory, it was created and studied focused on decision theory and its applications, without much concern with statistical formalisms. Thus, the objective of this work was to study important inferential concepts, such as point estimation and hypothesis testing for the Bernoulli model in presence of adversaries. We illustrate how these changes impact the point estimate and the Bayesian hypothesis test, besides the distribution of observable data and important statistical elements such as the behavior of Bayesian risk and critical regions.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Dissertacao.pdf (926.31 Kbytes)
Publishing Date
2021-07-02
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.