Distilling Intractable Generative Models

George Papamakarios; Iain Murray

Distilling Intractable Generative Models

George Papamakarios, Iain Murray

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

A generative model’s partition function is typically expressed as an intractable multi-dimensional integral, whose approximation presents a challenge to numerical and Monte Carlo integration. In this work, we propose a new estimation method for intractable partition functions, based on distilling an intractable generative model into a tractable approximation thereof, and using the latter for proposing Monte Carlo samples. We empirically demonstrate that our method -of-the-art estimates, even in combination with simple Monte Carlo methods.

Original language	English
Title of host publication	Probabilistic Integration Workshop at the Neural Information Processing Systems Conference, 2015
Number of pages	5
Publication status	Accepted/In press - 2 Aug 2015

Access to Document

distilling_intractable_generative_models_camera_readyAccepted author manuscript, 152 KB

Cite this

@inproceedings{8c14e8dabdfe4a7da84cb401a48533f3,

title = "Distilling Intractable Generative Models",

abstract = "A generative model{\textquoteright}s partition function is typically expressed as an intractable multi-dimensional integral, whose approximation presents a challenge to numerical and Monte Carlo integration. In this work, we propose a new estimation method for intractable partition functions, based on distilling an intractable generative model into a tractable approximation thereof, and using the latter for proposing Monte Carlo samples. We empirically demonstrate that our method -of-the-art estimates, even in combination with simple Monte Carlo methods.",

author = "George Papamakarios and Iain Murray",

year = "2015",

month = aug,

day = "2",

language = "English",

booktitle = "Probabilistic Integration Workshop at the Neural Information Processing Systems Conference, 2015",

}

TY - GEN

T1 - Distilling Intractable Generative Models

AU - Papamakarios, George

AU - Murray, Iain

PY - 2015/8/2

Y1 - 2015/8/2

N2 - A generative model’s partition function is typically expressed as an intractable multi-dimensional integral, whose approximation presents a challenge to numerical and Monte Carlo integration. In this work, we propose a new estimation method for intractable partition functions, based on distilling an intractable generative model into a tractable approximation thereof, and using the latter for proposing Monte Carlo samples. We empirically demonstrate that our method -of-the-art estimates, even in combination with simple Monte Carlo methods.

AB - A generative model’s partition function is typically expressed as an intractable multi-dimensional integral, whose approximation presents a challenge to numerical and Monte Carlo integration. In this work, we propose a new estimation method for intractable partition functions, based on distilling an intractable generative model into a tractable approximation thereof, and using the latter for proposing Monte Carlo samples. We empirically demonstrate that our method -of-the-art estimates, even in combination with simple Monte Carlo methods.

M3 - Conference contribution

BT - Probabilistic Integration Workshop at the Neural Information Processing Systems Conference, 2015

ER -

Distilling Intractable Generative Models

Abstract

Access to Document

Fingerprint

Cite this