Sampling-free Uncertainty Estimation in Gated Recurrent Units with Applications to Normative Modeling in Neuroimaging

Seong Jae Hwang, Ronak R. Mehta, Hyunwoo J. Kim, Sterling C. Johnson, Vikas Singh

Research output: Contribution to conferencePaper

Abstract

There has recently been a concerted effort to derive mechanisms in vision and machine learning systems to offer uncertainty estimates of the predictions they make. Clearly, there are benefits to a system that is not only accurate but also has a sense for when it is not. Existing proposals center around Bayesian interpretations of modern deep architectures – these are effective but can often be computationally demanding. We show how classical ideas in the literature on exponential families on probabilistic networks provide an excellent starting point to derive uncertainty estimates in Gated Recurrent Units (GRU). Our proposal directly quantifies uncertainty deterministically, without the need for costly sampling-based estimation. We show that while uncertainty is quite useful by itself in computer vision and machine learning, we also demonstrate that it can play a key role in enabling statistical analysis with deep networks in neuroimaging studies with normative modeling methods. To our knowledge, this is the first result describing sampling-free uncertainty estimation for powerful sequential models such as GRUs.

Original languageEnglish
Publication statusPublished - 2019 Jan 1
Event35th Conference on Uncertainty in Artificial Intelligence, UAI 2019 - Tel Aviv, Israel
Duration: 2019 Jul 222019 Jul 25

Conference

Conference35th Conference on Uncertainty in Artificial Intelligence, UAI 2019
CountryIsrael
CityTel Aviv
Period19/7/2219/7/25

Fingerprint

Neuroimaging
Sampling
Learning systems
Computer vision
Statistical methods
Uncertainty

ASJC Scopus subject areas

  • Artificial Intelligence

Cite this

Hwang, S. J., Mehta, R. R., Kim, H. J., Johnson, S. C., & Singh, V. (2019). Sampling-free Uncertainty Estimation in Gated Recurrent Units with Applications to Normative Modeling in Neuroimaging. Paper presented at 35th Conference on Uncertainty in Artificial Intelligence, UAI 2019, Tel Aviv, Israel.

Sampling-free Uncertainty Estimation in Gated Recurrent Units with Applications to Normative Modeling in Neuroimaging. / Hwang, Seong Jae; Mehta, Ronak R.; Kim, Hyunwoo J.; Johnson, Sterling C.; Singh, Vikas.

2019. Paper presented at 35th Conference on Uncertainty in Artificial Intelligence, UAI 2019, Tel Aviv, Israel.

Research output: Contribution to conferencePaper

Hwang, SJ, Mehta, RR, Kim, HJ, Johnson, SC & Singh, V 2019, 'Sampling-free Uncertainty Estimation in Gated Recurrent Units with Applications to Normative Modeling in Neuroimaging', Paper presented at 35th Conference on Uncertainty in Artificial Intelligence, UAI 2019, Tel Aviv, Israel, 19/7/22 - 19/7/25.
Hwang SJ, Mehta RR, Kim HJ, Johnson SC, Singh V. Sampling-free Uncertainty Estimation in Gated Recurrent Units with Applications to Normative Modeling in Neuroimaging. 2019. Paper presented at 35th Conference on Uncertainty in Artificial Intelligence, UAI 2019, Tel Aviv, Israel.
Hwang, Seong Jae ; Mehta, Ronak R. ; Kim, Hyunwoo J. ; Johnson, Sterling C. ; Singh, Vikas. / Sampling-free Uncertainty Estimation in Gated Recurrent Units with Applications to Normative Modeling in Neuroimaging. Paper presented at 35th Conference on Uncertainty in Artificial Intelligence, UAI 2019, Tel Aviv, Israel.
@conference{2a15186f81f3454b98298628e000548e,
title = "Sampling-free Uncertainty Estimation in Gated Recurrent Units with Applications to Normative Modeling in Neuroimaging",
abstract = "There has recently been a concerted effort to derive mechanisms in vision and machine learning systems to offer uncertainty estimates of the predictions they make. Clearly, there are benefits to a system that is not only accurate but also has a sense for when it is not. Existing proposals center around Bayesian interpretations of modern deep architectures – these are effective but can often be computationally demanding. We show how classical ideas in the literature on exponential families on probabilistic networks provide an excellent starting point to derive uncertainty estimates in Gated Recurrent Units (GRU). Our proposal directly quantifies uncertainty deterministically, without the need for costly sampling-based estimation. We show that while uncertainty is quite useful by itself in computer vision and machine learning, we also demonstrate that it can play a key role in enabling statistical analysis with deep networks in neuroimaging studies with normative modeling methods. To our knowledge, this is the first result describing sampling-free uncertainty estimation for powerful sequential models such as GRUs.",
author = "Hwang, {Seong Jae} and Mehta, {Ronak R.} and Kim, {Hyunwoo J.} and Johnson, {Sterling C.} and Vikas Singh",
year = "2019",
month = "1",
day = "1",
language = "English",
note = "35th Conference on Uncertainty in Artificial Intelligence, UAI 2019 ; Conference date: 22-07-2019 Through 25-07-2019",

}

TY - CONF

T1 - Sampling-free Uncertainty Estimation in Gated Recurrent Units with Applications to Normative Modeling in Neuroimaging

AU - Hwang, Seong Jae

AU - Mehta, Ronak R.

AU - Kim, Hyunwoo J.

AU - Johnson, Sterling C.

AU - Singh, Vikas

PY - 2019/1/1

Y1 - 2019/1/1

N2 - There has recently been a concerted effort to derive mechanisms in vision and machine learning systems to offer uncertainty estimates of the predictions they make. Clearly, there are benefits to a system that is not only accurate but also has a sense for when it is not. Existing proposals center around Bayesian interpretations of modern deep architectures – these are effective but can often be computationally demanding. We show how classical ideas in the literature on exponential families on probabilistic networks provide an excellent starting point to derive uncertainty estimates in Gated Recurrent Units (GRU). Our proposal directly quantifies uncertainty deterministically, without the need for costly sampling-based estimation. We show that while uncertainty is quite useful by itself in computer vision and machine learning, we also demonstrate that it can play a key role in enabling statistical analysis with deep networks in neuroimaging studies with normative modeling methods. To our knowledge, this is the first result describing sampling-free uncertainty estimation for powerful sequential models such as GRUs.

AB - There has recently been a concerted effort to derive mechanisms in vision and machine learning systems to offer uncertainty estimates of the predictions they make. Clearly, there are benefits to a system that is not only accurate but also has a sense for when it is not. Existing proposals center around Bayesian interpretations of modern deep architectures – these are effective but can often be computationally demanding. We show how classical ideas in the literature on exponential families on probabilistic networks provide an excellent starting point to derive uncertainty estimates in Gated Recurrent Units (GRU). Our proposal directly quantifies uncertainty deterministically, without the need for costly sampling-based estimation. We show that while uncertainty is quite useful by itself in computer vision and machine learning, we also demonstrate that it can play a key role in enabling statistical analysis with deep networks in neuroimaging studies with normative modeling methods. To our knowledge, this is the first result describing sampling-free uncertainty estimation for powerful sequential models such as GRUs.

UR - http://www.scopus.com/inward/record.url?scp=85073216551&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85073216551&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:85073216551

ER -