Visualisation of high-dimensional data using an ensemble of neural networks

Gianniotis, Nikolaos; Riggelsen, Carsten

doi:10.1109/ciel.2013.6613135

Cited by 3 publications

(3 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our assumption is that the N D -dimensional space is then projected to two dimensions, enabling visualization. There are several different methods to project highdimensional data onto a lower dimensional space, such as those described in Hotelling (1933), Gianniotis andRiggelsen (2013), andLawrence (2004). NGA-East used Sammon's mapping (Sammon, 1969), which is a relatively simple method that works on a continuous scale.…”

Section: Nga-east Implementation With the Covariance Model Developed Inmentioning

confidence: 99%

NGA-East Ground-Motion Characterization model part I: Summary of products and model development

et al. 2021

View full text Add to dashboard Cite

In this article, we present an overview of the research project NGA-East, Next Generation Attenuation for Central and Eastern North America (CENA), and summarize the key methodology and products. The project was tasked with developing a new ground motion characterization (GMC) model for CENA. The final NGA-East GMC model includes a set of 17 median ground motion models (GMMs) for peak ground acceleration and velocity (PGA, PGV) and response spectral ordinates for periods ranging from 0.01 to 10 s. The NGA-East GMMs are applicable to horizontal components of ground motions on very hard rock, for the moment magnitude range of 4.0–8.2, and distances of up to 1500 km. The aleatory standard deviations of GMMs are also provided for site-specific analysis (single-station standard deviation) and for general probabilistic seismic hazard analyses (PSHA) applications (ergodic standard deviation). In addition, adjustment factors are provided for source depth and hanging-wall effects, as well as for hazard computations at sites in the Gulf Coast Region. During the course of the project, several innovative technologies were developed and implemented to increase the transparency and repeatability of the GMC building process. This involved expanding on a set of candidate median GMMs to define and capture an appropriate range of epistemic uncertainty in ground motions. We also developed a new approach for modeling the aleatory variability that was completely independent of the median GMMs. The development made extensive use of the CENA database but also borrowed data from other parts of the world when relevant and led to an integrated suite of models. Through this repeatable process, epistemic uncertainty could be quantified more objectively than before, relying less on expert opinion. The NGA-East project went through a comprehensive Seismic Senior Hazard Analysis Committee (SSHAC) Level 3 peer review process before its release.

show abstract

Section: Nga-east Implementation With the Covariance Model Developed Inmentioning

confidence: 99%

NGA-East Ground-Motion Characterization model part I: Summary of products and model development

et al. 2021

View full text Add to dashboard Cite

show abstract

“…There are several different methods to project high-dimensional data onto a lowerdimensional space, such as those such as described in Hotelling [1933], Gianniotis and Riggelsen [2013], and Lawrence [2004]. Scherbaum et al [2010] used SOMs [Kohonen 2001] and Sammon's mapping [1969].…”

Section: Challenges In Evaluation Of Multiple Gmmsmentioning

confidence: 99%

NGA-East Ground-Motion Models for the U.S. Geological Survey National Seismic Hazard Maps

Goulet,

Bozorgnia,

Kuehn

et al. 2017

PEER Reports

View full text Add to dashboard Cite

The purpose of this report is to provide a set of ground motion models (GMMs) to be considered by the U.S. Geological Survey (USGS) for their National Seismic Hazard Maps (NSHMs) for the Central and Eastern U.S. (CEUS). These interim GMMs are adjusted and modified from a set of preliminary models developed as part of the Next Generation Attenuation for Central and Eastern North-America (CENA) project (NGA-East). The NGA-East objective was to develop a new ground-motion characterization (GMC) model for the CENA region. The GMC model consists of a set of GMMs for median and standard deviation of ground motions and their associated weights in the logic-tree for use in probabilistic seismic hazard analysis (PSHA). NGA-East is a large multidisciplinary project coordinated by the Pacific Earthquake Engineering Research Center (PEER), at the University of California, Berkeley. The project has two components: (1) a set of scientific research tasks, and (2) a model-building component following the framework of the “Seismic Senior Hazard Analysis Committee (SSHAC) Level 3” [Budnitz et al. 1997; NRC 2012]. Component (2) is built on the scientific results of component (1) of the NGA-East Project. This report does not document the final NGA-East model under (2), but instead presents interim GMMs for use in the U.S. Geological Survey (USGS) National Seismic Hazard Maps. Under component (1) of NGA-East, several scientific issues were addressed, including: (a) development of a new database of empirical data recorded in CENA; (b) development of a regionalized ground-motion map for CENA, (c) definition of the reference site condition; (d) simulations of ground motions based on different methodologies, (e) development of numerous GMMs for CENA, and (f) the development of the current report. The scientific tasks of NGA- East were all documented as a series of PEER reports. This report documents the GMMs recommended by the authors for consideration by the USGS for their NSHM. The report documents the key elements involved in the development of the proposed GMMs and summarizes the median and aleatory models for ground motions along with their recommended weights. The models presented here build on the work from the authors and aim to globally represent the epistemic uncertainty in ground motions for CENA. The NGA-East models for the USGS NSHMs includes a set of 13 GMMs defined for 25 ground-motion intensity measures, applicable to CENA in the moment magnitude range of 4.0 to 8.2 and covering distances up to 1500 km. Standard deviation models are also provided for general PSHA applications (ergodic standard deviation). Adjustment factors are provided for hazard computations involving the Gulf Coast region.

show abstract

“…For example, Ye et al [3] introduced a -nearest neighbor based bagging algorithm with pruning for ensemble SVM classification. Ensembles have also proven well in other applications like visualization of high-dimensional data with neural networks [4].…”

Section: Introductionmentioning

confidence: 99%

Cascade Support Vector Machines with Dimensionality Reduction

Krämer

2015

Applied Computational Intelligence and Soft Computing

View full text Add to dashboard Cite

Cascade support vector machines have been introduced as extension of classic support vector machines that allow a fast training on large data sets. In this work, we combine cascade support vector machines with dimensionality reduction based preprocessing. The cascade principle allows fast learning based on the division of the training set into subsets and the union of cascade learning results based on support vectors in each cascade level. The combination with dimensionality reduction as preprocessing results in a significant speedup, often without loss of classifier accuracies, while considering the high-dimensional pendants of the low-dimensional support vectors in each new cascade level. We analyze and compare various instantiations of dimensionality reduction preprocessing and cascade SVMs with principal component analysis, locally linear embedding, and isometric mapping. The experimental analysis on various artificial and real-world benchmark problems includes various cascade specific parameters like intermediate training set sizes and dimensionalities.

show abstract

Visualisation of high-dimensional data using an ensemble of neural networks

Cited by 3 publications

References 18 publications

NGA-East Ground-Motion Characterization model part I: Summary of products and model development

NGA-East Ground-Motion Characterization model part I: Summary of products and model development

NGA-East Ground-Motion Models for the U.S. Geological Survey National Seismic Hazard Maps

Cascade Support Vector Machines with Dimensionality Reduction

Contact Info

Product

Resources

About