Semi-supervised Bayesian Low-shot Learning

Adams, Jason; Goode, Katherine; Michalenko, Joshua; Lewis, Phillip; Ries, Daniel; Zollweg, Josh

doi:10.2172/1821543

Title: Semi-supervised Bayesian Low-shot Learning

Full Record
Other Related Research

Abstract

Deep neural networks (NNs) typically outperform traditional machine learning (ML) approaches for complicated, non-linear tasks. It is expected that deep learning (DL) should offer superior performance for the important non-proliferation task of predicting explosive device configuration based upon observed optical signature, a task which human experts struggle with. However, supervised machine learning is difficult to apply in this mission space because most recorded signatures are not associated with the corresponding device description, or “truth labels.” This is challenging for NNs, which traditionally require many samples for strong performance. Semi-supervised learning (SSL), low-shot learning (LSL), and uncertainty quantification (UQ) for NNs are emerging approaches that could bridge the mission gaps of few labels and rare samples of importance. NN explainability techniques are important in gaining insight into the inferential feature importance of such a complex model. In this work, SSL, LSL, and UQ are merged into a single framework, a significant technical hurdle not previously demonstrated. Exponential Average Adversarial Training (EAAT) and Pairwise Neural Networks (PNNs) are chosen as the SSL and LSL methods of choice. Permutation feature importance (PFI) for functional data is used to provide explainability via the Variable importance Explainable Elastic Shape Analysis (VEESA) pipeline. A variety ofmore »« less

Authors:

Adams, Jason ^[1]; Goode, Katherine ^[1]; Michalenko, Joshua ^[1]; Lewis, Phillip ^[1]; Ries, Daniel ^[1]; Zollweg, Josh ^[1]

Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)

Publication Date:: Wed Sep 01 00:00:00 EDT 2021

Research Org.:: Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)

Sponsoring Org.:: USDOE National Nuclear Security Administration (NNSA)

OSTI Identifier:: 1821543

Report Number(s):: SAND2021-11319
699613; TRN: US2301652

DOE Contract Number:: NA0003525

Resource Type:: Technical Report

Country of Publication:: United States

Language:: English

Subject:: 97 MATHEMATICS AND COMPUTING; 98 NUCLEAR DISARMAMENT, SAFEGUARDS, AND PHYSICAL PROTECTION

Citation Formats


                    Adams, Jason, Goode, Katherine, Michalenko, Joshua, Lewis, Phillip, Ries, Daniel, and Zollweg, Josh. Semi-supervised Bayesian Low-shot Learning.  United States: N. p., 2021. 
        Web.  doi:10.2172/1821543.

Copy to clipboard


                    Adams, Jason, Goode, Katherine, Michalenko, Joshua, Lewis, Phillip, Ries, Daniel, & Zollweg, Josh. Semi-supervised Bayesian Low-shot Learning.  United States.  https://doi.org/10.2172/1821543

Copy to clipboard


                    Adams, Jason, Goode, Katherine, Michalenko, Joshua, Lewis, Phillip, Ries, Daniel, and Zollweg, Josh. 2021.  
        "Semi-supervised Bayesian Low-shot Learning".  United States.  https://doi.org/10.2172/1821543.  https://www.osti.gov/servlets/purl/1821543.

Copy to clipboard


                    
@article{osti_1821543,

  title        = {Semi-supervised Bayesian Low-shot Learning},

  author       = {Adams, Jason and Goode, Katherine and Michalenko, Joshua and Lewis, Phillip and Ries, Daniel and Zollweg, Josh},

  abstractNote = {Deep neural networks (NNs) typically outperform traditional machine learning (ML) approaches for complicated, non-linear tasks. It is expected that deep learning (DL) should offer superior performance for the important non-proliferation task of predicting explosive device configuration based upon observed optical signature, a task which human experts struggle with. However, supervised machine learning is difficult to apply in this mission space because most recorded signatures are not associated with the corresponding device description, or “truth labels.” This is challenging for NNs, which traditionally require many samples for strong performance. Semi-supervised learning (SSL), low-shot learning (LSL), and uncertainty quantification (UQ) for NNs are emerging approaches that could bridge the mission gaps of few labels and rare samples of importance. NN explainability techniques are important in gaining insight into the inferential feature importance of such a complex model. In this work, SSL, LSL, and UQ are merged into a single framework, a significant technical hurdle not previously demonstrated. Exponential Average Adversarial Training (EAAT) and Pairwise Neural Networks (PNNs) are chosen as the SSL and LSL methods of choice. Permutation feature importance (PFI) for functional data is used to provide explainability via the Variable importance Explainable Elastic Shape Analysis (VEESA) pipeline. A variety of uncertainty quantification approaches are explored: Bayesian Neural Networks (BNNs), ensemble methods, concrete dropout, and evidential deep learning. Two final approaches, one utilizing ensemble methods and one utilizing evidential learning, are constructed and compared using a well-quantified synthetic 2D dataset along with the DIRSIG Megascene.},

  doi          = {10.2172/1821543},

  url          = {https://www.osti.gov/biblio/1821543},
  journal      = {},
number       = ,

  volume       = ,

  place        = {United States},

  year         = {Wed Sep 01 00:00:00 EDT 2021},

  month        = {Wed Sep 01 00:00:00 EDT 2021}

}

Copy to clipboard

Technical Report:

View Technical Report (22.07 MB)

https://doi.org/10.2172/1821543

Save / Share:

Export Metadata

Save to My Library

Similar records in OSTI.GOV collections:

Final Technical Report - Applications of Machine Learning Techniques to Geothermal Play Fairway Analysis in the Great Basin Region, Nevada

Technical Report Faulds, James; Smith, Connor; Brown, Stephen; ...

Geothermal play fairway analysis (PFA) is a methodology that integrates geological, geophysical, and geochemical parameters to produce geothermal potential maps and identify promising areas for hidden systems. Although a previous PFA was successful in identifying new geothermal systems in Nevada, key challenges affected the PFA, including the estimation of parameter weights, a limited number of training sites, and limitations on some datasets. In this project, we applied machine learning (ML) techniques, including both supervised and unsupervised approaches, to ~1/3 of Nevada to mitigate those limitations and better characterize signatures for subsurface permeability related to hidden systems. Major project activities andmore »« less
https://doi.org/10.2172/2335471

Full Text Available
Deeply uncertain: comparing methods of uncertainty quantification in deep learning algorithms

Journal Article Caldeira, João; Nord, Brian - Machine Learning: Science and Technology

We present a comparison of methods for uncertainty quantification (UQ) in deep learning algorithms in the context of a simple physical system. Three of the most common uncertainty quantification methods - Bayesian Neural Networks (BNN), Concrete Dropout (CD), and Deep Ensembles (DE) - are compared to the standard analytic error propagation. We discuss this comparison in terms endemic to both machine learning ("epistemic" and "aleatoric") and the physical sciences ("statistical" and "systematic"). The comparisons are presented in terms of simulated experimental measurements of a single pendulum - a prototypical physical system for studying measurement and analysis techniques. Our results highlightmore »« less
https://doi.org/10.1088/2632-2153/aba6f3
Semisupervised Learning for Seismic Monitoring Applications

Journal Article Linville, Lisa; Anderson, Dylan; Michalenko, Joshua; ... - Seismological Research Letters

The impressive performance that deep neural networks demonstrate on a range of seismic monitoring tasks depends largely on the availability of event catalogs that have been manually curated over many years or decades. However, the quality, duration, and availability of seismic event catalogs vary significantly across the range of monitoring operations, regions, and objectives. Semisupervised learning (SSL) enables learning from both labeled and unlabeled data and provides a framework to leverage the abundance of unreviewed seismic data for training deep neural networks on a variety of target tasks. We apply two SSL algorithms (mean-teacher and virtual adversarial training) as wellmore »« less
https://doi.org/10.1785/0220200195

Full Text Available
Machine Learning Assisted Safety Modeling and Analysis of Advanced Reactors

Technical Report Liu, Yang; Hu, Rui; Dai, Dezhi; ...

With the advances in computational power and numerical methods, analysts can now rely on first-principle simulations to predict ultra-fine details in a variety of applications. Advances in machine learning (ML) have produced algorithms that can now learn high-level abstractions via hierarchical models. This project aims to leverage advances in ML techniques and the available high-resolution simulation data to develop a novel modeling and simulation (M\&S) methodology for reactor safety analysis. While application-agnostic ML techniques are available, complex physics constraints need to be incorporated into ML techniques to build ML-based closures for computationally efficient predictive simulations. This project intends to developmore »« less
https://doi.org/10.2172/1838607

Full Text Available
We Know Where We Don't Know: 3D Bayesian CNNs for Credible Geometric Uncertainty

Technical Report LaBonte, Tyler; Martinez, Carianne; Roberts, Scott

Deep learning has been successfully applied to the segmentation of 3D Computed Tomography (CT) scans. Establishing the credibility of these segmentations requires uncertainty quantification (UQ) to identify untrustworthy predictions. Recent UQ architectures include Monte Carlo dropout networks (MCDNs), which approximate deep Gaussian processes, and Bayesian neural networks (BNNs), which learn the distribution of the weight space. BNNs are advantageous over MCDNs for UQ but are thought to be computationally infeasible in high dimension, and neither architecture has produced interpretable geometric uncertainty maps. We propose a novel 3D Bayesian convolutional neural network (BCNN), the first deep learning method which generates statisticallymore »« less
https://doi.org/10.2172/1605518

Full Text Available

Similar Records