Boosting Black-Box Variational Inference by Incorporating the Natural Gradient

Felix  Trusheim; Alexandru Paul Condurache; Alfred Mertins

Boosting Black-Box Variational Inference by Incorporating the Natural Gradient

Felix Trusheim, Alexandru Paul Condurache, Alfred Mertins

Institute of Signal Processing

Abstract

In this paper we present a modification of thepopular Black-Box Variational Inference (BBVI) approachwhich significantly improves the computational efficiency of theinference. We achieve this performance boost by replacing thestandard gradient in the stochastic gradient ascent framework ofBBVI with the natural gradient. Our experimental results (e.g.training of neutral networks) show that the proposed methodoutperforms the original BBVI algorithm on both synthetic andreal data.

Original language	English
Number of pages	6
Publication status	Published - 01.08.2018

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 9 Industry, Innovation, and Infrastructure

Access to Document

https://www.isip.uni-luebeck.de/fileadmin/files/Trusheim2018.pdf

Cite this

@conference{b68ffd73f8e14ceea5cdcce6e88b2654,

title = "Boosting Black-Box Variational Inference by Incorporating the Natural Gradient",

abstract = "In this paper we present a modification of thepopular Black-Box Variational Inference (BBVI) approachwhich significantly improves the computational efficiency of theinference. We achieve this performance boost by replacing thestandard gradient in the stochastic gradient ascent framework ofBBVI with the natural gradient. Our experimental results (e.g.training of neutral networks) show that the proposed methodoutperforms the original BBVI algorithm on both synthetic andreal data.",

author = "Felix Trusheim and Condurache, \{Alexandru Paul\} and Alfred Mertins",

year = "2018",

month = aug,

day = "1",

language = "English",

}

TY - CONF

T1 - Boosting Black-Box Variational Inference by Incorporating the Natural Gradient

AU - Trusheim, Felix

AU - Condurache, Alexandru Paul

AU - Mertins, Alfred

PY - 2018/8/1

Y1 - 2018/8/1

N2 - In this paper we present a modification of thepopular Black-Box Variational Inference (BBVI) approachwhich significantly improves the computational efficiency of theinference. We achieve this performance boost by replacing thestandard gradient in the stochastic gradient ascent framework ofBBVI with the natural gradient. Our experimental results (e.g.training of neutral networks) show that the proposed methodoutperforms the original BBVI algorithm on both synthetic andreal data.

AB - In this paper we present a modification of thepopular Black-Box Variational Inference (BBVI) approachwhich significantly improves the computational efficiency of theinference. We achieve this performance boost by replacing thestandard gradient in the stochastic gradient ascent framework ofBBVI with the natural gradient. Our experimental results (e.g.training of neutral networks) show that the proposed methodoutperforms the original BBVI algorithm on both synthetic andreal data.

M3 - Conference Papers

ER -

Boosting Black-Box Variational Inference by Incorporating the Natural Gradient

Abstract

UN SDGs

Access to Document

Fingerprint

Cite this