Orthogonal Approximation of Marginal Likelihood of Generative Models

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21230%2F19%3A00339857" target="_blank" >RIV/68407700:21230/19:00339857 - isvavai.cz</a>

Nalezeny alternativní kódy

RIV/67985556:_____/19:00522204

Výsledek na webu

<a href="http://bayesiandeeplearning.org/2019/papers/48.pdf" target="_blank" >http://bayesiandeeplearning.org/2019/papers/48.pdf</a>

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Orthogonal Approximation of Marginal Likelihood of Generative Models

Popis výsledku v původním jazyce

This paper presents a new approximation of the marginal likelihood of generativemodels which is used as a score for anomaly detection. The score is motivatedby the shortcoming of the popular reconstruction error that it can behave arbitrar-ily outside the known samples. The proposed score corrects this by orthogonalcombination of the reconstruction error and the likelihood in the latent space. Asexperimentally shown on benchmark problems from anomaly detection and illus-trated on a toy problem, this combination lends the score robustness to outliers.Generative models evaluated with this score outperformed the competing meth-ods especially in tasks of learning distribution from data corrupted by anomalies.Finally, the score is compatible with contemporary generative models, namelyvariational auto-encoders and generative adversarial networks.

Název v anglickém jazyce

Orthogonal Approximation of Marginal Likelihood of Generative Models

Popis výsledku anglicky

This paper presents a new approximation of the marginal likelihood of generativemodels which is used as a score for anomaly detection. The score is motivatedby the shortcoming of the popular reconstruction error that it can behave arbitrar-ily outside the known samples. The proposed score corrects this by orthogonalcombination of the reconstruction error and the likelihood in the latent space. Asexperimentally shown on benchmark problems from anomaly detection and illus-trated on a toy problem, this combination lends the score robustness to outliers.Generative models evaluated with this score outperformed the competing meth-ods especially in tasks of learning distribution from data corrupted by anomalies.Finally, the score is compatible with contemporary generative models, namelyvariational auto-encoders and generative adversarial networks.

Druh

O - Ostatní výsledky

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

Výsledek vznikl pri realizaci vícero projektů. Více informací v záložce Projekty.

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2019

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů