Search NASA⌕ Search

SEARCH · Search NASA

Results for “risk classification”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 163 records · Page 9
All sourcesNASA NTRSDOE OSTI
DOE OSTI2018

Investigating Sociodemographic Disparities in Cancer Risk Using Web-Based Informatics

Cancer health disparities due to demographic and socioeconomic factors are an area of great interest in the epidemiological community. Adjusting for such factors is important when developing cancer risk models. However, for digital epidemiology studies relying on online sources such information is not readily available. This paper presents a novel method for extracting demographic and socioeconomic information from openly available online obituaries. The method relies on tailored language processing rules and a probabilistic scheme to map subjects’ occupation history to the occupation classification codes and related earnings provided by the U.S. Census Bureau. Using this information, a case-control study is executed fully in silico to investigate how age, gender, parity, and income level impact breast and lung cancer risk. Based on 48,368 online obituaries (4,643 for breast cancer, 6,274 for lung cancer, and 37,451 cancer-free) collected automatically and a generalized cancer risk model, our study shows strong association between age, parity, and socioeconomic status and cancer risk. Although for breast cancer the observed trends are very consistent with traditional epidemiological studies, some inconsistency is observed for lung cancer with respect to socioeconomic status.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2024

Root Cause Correlation Analysis of Software Failures via Orthogonal Defect Classification and Natural Language Processing

Systems theoretic process analysis (STPA) is becoming an increasingly popular technique to assess how complex digital software systems can fail. Rather than defining failures by their observable failure events, which may be sparse especially for safety rated nuclear digital instrumentation and control systems (DI&C), failures are defined as postulated unsafe actions under specific contextual conditions. This permits a top-down analysis of system hazards and identifies whether imposed constraints and requirements can sufficiently address undesirable hazards. However, STPA is a qualitative approach at identifying inadequacies in the development process and cannot currently be used to quantify unsafe action likelihoods for probabilistic risk assessment. Therefore, in this work, we examine the root causes of software failure and explore whether a consistent correlation can be linked to specific unsafe action classes. We implement Lbl2Vec, an unsupervised document classification and retrieval algorithm, on a database of 4,096 software defect reports acquired from various open-source software systems. By analyzing sentence structure, embedded labels, and word vectors, we show that certain defect types positively correlate to specific unsafe action classes over others. The correlations developed can be used to estimate the failure probability of safety intended DI&C systems which provides a licensing basis for nuclear plant modernization efforts.

22 - GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2024

Root Cause Correlation Analysis of Software Failures via Orthogonal Defect Classification and Natural Language Processing

Systems theoretic process analysis (STPA) is becoming an increasingly popular technique to assess how complex digital software systems can fail. Rather than defining failures by their observable failure events, which may be sparse especially for safety rated nuclear digital instrumentation and control systems (DI&C), failures are defined as postulated unsafe actions under specific contextual conditions. This permits a top-down analysis of system hazards and identifies whether imposed constraints and requirements can sufficiently address undesirable hazards. However, STPA is a qualitative approach at identifying inadequacies in the development process and cannot currently be used to quantify unsafe action likelihoods for probabilistic risk assessment. Therefore, in this work, we examine the root causes of software failure and explore whether a consistent correlation can be linked to specific unsafe action classes. We implement Lbl2Vec, an unsupervised document classification and retrieval algorithm, on a database of 4,096 software defect reports acquired from various open-source software systems. By analyzing sentence structure, embedded labels, and word vectors, we show that certain defect types positively correlate to specific unsafe action classes over others. The correlations developed can be used to estimate the failure probability of safety intended DI&C systems which provides a licensing basis for nuclear plant modernization efforts.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2015

Chooz-A steam generators characterization - 15205

EDF nuclear waste management requires a deep understanding of characterization, classification and waste sorting operations. In fact, French nuclear waste management defines several classes with specific management, treatment and storage facilities. Based on particular criteria, the more the radiological risk of the nuclear waste is important, the more its management will be complex and expensive. During the dismantling of the first French pressurized reactor Chooz-A, decontamination of the primary water circuits (not including the reactor vessel), the steam generators and the pressurizer have been carried out in order to reduce their activity levels. Thanks to these decontamination operations, and a specific characterization methodology, EDF was able to de-classify the 4 steam generators and store them in one piece at the ANDRA Very Low Level Activity disposal facilities instead of the Low and Intermediate Level activity one. This de-classification allowed EDF to avoid important cutting and packaging processes. Therefore, significant dose reductions for the operators and financial gain have been made. To characterize and declare Chooz-A SG activity, EDF-CIDEN uses a methodology defined by the French institute of atomic energy, CEA. The method is based on external gamma spectrometry measurements performed with NaI collimated detectors, associated with MERCURAD simulations providing the transfer functions for the detectors and activity sources. Internal measurements are carried out with a CZT (CdZnTe) probe inside the SG tubes to refine the 3D model. In fact, the primary side represents the main source of activity, and understanding its contamination distribution is important to reduce the model and calculation uncertainties. Measurements eventually provide SG {sup 60}Co global activity, from which the activities of other radionuclides of the spectrum are determined using scaling factors. The final activity declaration takes into account the standard deviation of the measurements in order to cover the uncertainties of the methodology. Thereby, the declaration is 'slightly over-estimative', enough to meet the waste storage facility acceptance criteria. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
NASA NTRS1975

Present and potential land use mapping in Mexico

The Mexican Water Plan (MWP) conducted studies of present and potential land use in Mexico using LANDSAT-1 satellite imagery. Present land use studies were carried out all over the country (197 million hectares); nine soil uses were mapped according to the first classification level recommended by the U.S. Geological Survey. Also 6.3 million hectares of land with advanced erosion were detected. Work was executed at a rate of 8 million hectares per month; reliability was 90% and the cost of only 0.1 cents/hectare. The potential land use study was performed in 45 million hectares at a rate of 4 million hectares per month and at a cost of 0.33 cents/hectare. Soil units according to FAO classification were delineated scale 1:1 million; interpretative maps were also prepared dealing with potential agricultural productivity carrying capacity for cattle, water, erosion risk, and slope ranges.

Garduno, H.↗
NASA NTRS1996

Advanced Refrigerator/Freezer Technology Development. Technology Assessment

The NASA Lewis Research Center, through contract with Oceaneering Space Systems, is engaged in a project to develop advanced refrigerator/freezer (R/F) technologies for future Life and Biomedical Sciences space flight missions. The first phase of this project, a technology assessment, has been completed to identify the advanced R/F technologies needed and best suited to meet the requirements for the five R/F classifications specified by Life and Biomedical Science researchers. Additional objectives of the technology assessment were to rank those technologies based on benefit and risk, and to recommend technology development activities that can be accomplished within this project. This report presents the basis, the methodology, and results of the R/F technology assessment, along with technology development recommendations.

Gaseor, Thomas↗
DOE OSTI2024

Question-answering system extracts information on injection drug use from clinical notes

Background. Injection drug use (IDU) can increase mortality and morbidity. Therefore, identifying IDU early and initiating harm reduction interventions can benefit individuals at risk. However, extracting IDU behaviors from patients’ electronic health records (EHR) is difficult because there is no other structured data available, such as International Classification of Disease (ICD) codes, and IDU is most often documented in unstructured free-text clinical notes. Although natural language processing can efficiently extract this information from unstructured data, there are no validated tools. Methods. Here, to address this gap in clinical information, we design a question-answering (QA) framework to extract information on IDU from clinical notes for use in clinical operations. Our framework involves two main steps: (1) generating a gold-standard QA dataset and (2) developing and testing the QA model. We use 2323 clinical notes of 1145 patients curated from the US Department of Veterans Affairs (VA) Corporate Data Warehouse to construct the gold-standard dataset for developing and evaluating the QA model. We also demonstrate the QA model’s ability to extract IDU-related information from temporally out-of-distribution data. Results. Here, we show that for a strict match between gold-standard and predicted answers, the QA model achieves a 51.65% F1 score. For a relaxed match between the gold-standard and predicted answers, the QA model obtains a 78.03% F1 score, along with 85.38% Precision and 79.02% Recall scores. Moreover, the QA model demonstrates consistent performance when subjected to temporally out-of-distribution data. Conclusions. Our study introduces a QA framework designed to extract IDU information from clinical notes, aiming to enhance the accurate and efficient detection of people who inject drugs, extract relevant information, and ultimately facilitate informed patient care.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2022

Utilization of the LMP Methodology in Support of the VTR Conceptual Safety Design Report

The Versatile Test Reactor (VTR) is a fast spectrum test reactor currently being developed in the United States under the direction of the US Department of Energy (DOE), Office of Nuclear Energy. The VTR is utilizing a risk-informed performance-based (RIPB) approach for design support and authorization by the DOE, derived from recent efforts by the US industry led Licensing Modernization Project (LMP). This document contains an overview of the implementation of the LMP approach in support of the VTR Conceptual Safety Design Report (CSDR). The work reported here is the result of studies supporting a VTR conceptual design, cost, and schedule estimate for DOE-NE to make a decision on procurement. As such, it is preliminary. The VTR RIPB authorization approach utilizes information from the probabilistic risk assessment (PRA), coupled with deterministic analyses, to aid in decision-making regarding the identification and categorization of safety basis events (SBEs), the classification of structures, systems, and components (SSCs), and the evaluation of defense-in-depth (DID) adequacy. As part of initial reactor design efforts, a VTR conceptual design PRA was developed to support the RIPB process, which focused on at-power internal events, with scoping analyses for seismic and sodium fire hazards. In addition to supporting numerous design studies, preliminary results from the RIPB approach and the VTR conceptual design PRA were utilized as the basis of the VTR CSDR. The initial identification and categorization of SBEs, SSC classification, and DID evaluation were contained within the CSDR, which was submitted to DOE in 2019 as part of the CD-1 submittal package. Following review, DOE approved the CSDR in April 2020 and the CD-1 package in late 2020. Valuable experience was gained through the implementation of the RIPB approach for design and authorization during the VTR conceptual design phase, which is summarized in this document. To the extent possible, this experience has been shared with the advanced reactor industry, through publications and participation in licensing tabletops, in addition to informing DOE:NE advanced reactor regulatory development efforts. Furthermore, the approval of the CSDR by the DOE as part of CD-1 represents a significant milestone in the use of RIPB approaches for advanced reactor licensing.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗
DOE OSTI2020

Cancer genomics predicts disease relapse and therapeutic response to neoadjuvant chemotherapy of hormone sensitive breast cancers

Several studies provide insight into the landscape of breast cancer genomics with the genomic characterization of tumors offering exceptional opportunities in defining therapies tailored to the patient’s specific need. However, translating genomic data into personalized treatment regimens has been hampered partly due to uncertainties in deviating from guideline based clinical protocols. Here we report a genomic approach to predict favorable outcome to treatment responses thus enabling personalized medicine in the selection of specific treatment regimens. The genomic data were divided into a training set of N = 835 cases and a validation set consisting of 1315 hormone sensitive, 634 triple negative breast cancer (TNBC) and 1365 breast cancer patients with information on neoadjuvant chemotherapy responses. Patients were selected by the following criteria: estrogen receptor (ER) status, lymph node invasion, recurrence free survival. The k-means classification algorithm delineated clusters with low- and high- expression of genes related to recurrence of disease; a multivariate Cox’s proportional hazard model defined recurrence risk for disease. Classifier genes were validated by Immunohistochemistry (IHC) using tissue microarray sections containing both normal and cancerous tissues and by evaluating findings deposited in the human protein atlas repository. Based on the leave-on-out cross validation procedure of 4 independent data sets we identified 51-genes associated with disease relapse and selected 10, i.e. TOP2A, AURKA, CKS2, CCNB2, CDK1 SLC19A1, E2F8, E2F1, PRC1, KIF11 for in depth validation. Expression of the mechanistically linked disease regulated genes significantly correlated with recurrence free survival among ER-positive and triple negative breast cancer patients and was independent of age, tumor size, histological grade and node status. Importantly, the classifier genes predicted pathological complete responses to neoadjuvant chemotherapy (P < 0.001) with high expression of these genes being associated with an improved therapeutic response toward two different anthracycline-taxane regimens; thus, highlighting the prospective for precision medicine. Our study demonstrates the potential of classifier genes to predict risk for disease relapse and treatment response to chemotherapies. The classifier genes enable rational selection of patients who benefit best from a given chemotherapy thus providing the best possible care. The findings encourage independent clinical validation.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Evidential Deep Learning: Enhancing Predictive Uncertainty Estimation for Earth System Science Applications

Abstract Robust quantification of predictive uncertainty is a critical addition needed for machine learning applied to weather and climate problems to improve the understanding of what is driving prediction sensitivity. Ensembles of machine learning models provide predictive uncertainty estimates in a conceptually simple way but require multiple models for training and prediction, increasing computational cost and latency. Parametric deep learning can estimate uncertainty with one model by predicting the parameters of a probability distribution but does not account for epistemic uncertainty. Evidential deep learning, a technique that extends parametric deep learning to higher-order distributions, can account for both aleatoric and epistemic uncertainties with one model. This study compares the uncertainty derived from evidential neural networks to that obtained from ensembles. Through applications of the classification of winter precipitation type and regression of surface-layer fluxes, we show evidential deep learning models attaining predictive accuracy rivaling standard methods while robustly quantifying both sources of uncertainty. We evaluate the uncertainty in terms of how well the predictions are calibrated and how well the uncertainty correlates with prediction error. Analyses of uncertainty in the context of the inputs reveal sensitivities to underlying meteorological processes, facilitating interpretation of the models. The conceptual simplicity, interpretability, and computational efficiency of evidential neural networks make them highly extensible, offering a promising approach for reliable and practical uncertainty quantification in Earth system science modeling. To encourage broader adoption of evidential deep learning, we have developed a new Python package, Machine Integration and Learning for Earth Systems (MILES) group Generalized Uncertainty for Earth System Science (GUESS) (MILES-GUESS) ( https://github.com/ai2es/miles-guess ), that enables users to train and evaluate both evidential and ensemble deep learning. Significance Statement This study demonstrates a new technique, evidential deep learning, for robust and computationally efficient uncertainty quantification in modeling the Earth system. The method integrates probabilistic principles into deep neural networks, enabling the estimation of both aleatoric uncertainty from noisy data and epistemic uncertainty from model limitations using a single model. Our analyses reveal how decomposing these uncertainties provides valuable insights into reliability, accuracy, and model shortcomings. We show that the approach can rival standard methods in classification and regression tasks within atmospheric science while offering practical advantages such as computational efficiency. With further advances, evidential networks have the potential to enhance risk assessment and decision-making across meteorology by improving uncertainty quantification, a longstanding challenge. This work establishes a strong foundation and motivation for the broader adoption of evidential learning, where properly quantifying uncertainties is critical yet lacking.

Schreck, John S.↗
DOE OSTI2022

Post-Combat-Injury Opioid Prescription and Alcohol Use Disorder in the Military

Previous studies have identified combat exposure and combat traumatic experience as problematic drinking risk factors. Increasing evidence suggests that opioid use increases the risk of alcohol use disorder. This study investigated the association between opioid prescription use after injury and (1) alcohol use disorder and (2) severity of alcohol use disorder among deployed military servicemembers. Deidentified health records data of 9,029 deployed servicemembers from a retrospective cohort study were analyzed. Data were randomly selected from the Department of Defense Trauma Registry and included servicemembers with combat injuries during deployment in Iraq or Afghanistan (2002–2016). Pharmacy records and International Classification of Diseases, Ninth and Tenth Revision diagnosis codes were used. Three groups were identified (no opioid prescription use, nonpersistent opioid prescription use, and persistent opioid prescription use) and were compared on the basis of alcohol use disorder risk using Cox proportional hazard models. Data analyses were performed in 2021. Of the 9,029 servicemembers with combat injury, 2,262 developed alcohol use disorder (1,322 developed severe alcohol use disorder). Compared with no opioid prescription use, increased alcohol use disorder risk was associated with persistent opioid prescription use, with a hazard ratio of 1.13 (95% CI=1.02, 1.26). After covariate adjustment, increased risk remained statistically significant (hazards ratio=1.24; 95% CI=1.10, 1.39). There was no significant difference in alcohol use disorder risk between no opioid prescription use and nonpersistent opioid prescription use. The risk of severe alcohol use disorder did not vary by opioid use among servicemembers with alcohol use disorder diagnosis. Furthermore, the findings of the study suggest that the incidence of alcohol use disorder was higher among injured servicemembers with persistent opioid prescription use than among those without opioid use. If replicated in prospective studies, the findings highlight the need for clinicians to consider the current and history of alcohol use of patients in initiating treatment involving opioids.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2015

Large scale validation of the M5L lung CAD on heterogeneous CT datasets

Purpose: M5L, a fully automated computer-aided detection (CAD) system for the detection and segmentation of lung nodules in thoracic computed tomography (CT), is presented and validated on several image datasets. Methods: M5L is the combination of two independent subsystems, based on the Channeler Ant Model as a segmentation tool [lung channeler ant model (lungCAM)] and on the voxel-based neural approach. The lungCAM was upgraded with a scan equalization module and a new procedure to recover the nodules connected to other lung structures; its classification module, which makes use of a feed-forward neural network, is based of a small number of features (13), so as to minimize the risk of lacking generalization, which could be possible given the large difference between the size of the training and testing datasets, which contain 94 and 1019 CTs, respectively. The lungCAM (standalone) and M5L (combined) performance was extensively tested on 1043 CT scans from three independent datasets, including a detailed analysis of the full Lung Image Database Consortium/Image Database Resource Initiative database, which is not yet found in literature. Results: The lungCAM and M5L performance is consistent across the databases, with a sensitivity of about 70% and 80%, respectively, at eight false positive findings per scan, despite the variable annotation criteria and acquisition and reconstruction conditions. A reduced sensitivity is found for subtle nodules and ground glass opacities (GGO) structures. A comparison with other CAD systems is also presented. Conclusions: The M5L performance on a large and heterogeneous dataset is stable and satisfactory, although the development of a dedicated module for GGOs detection could further improve it, as well as an iterative optimization of the training procedure. The main aim of the present study was accomplished: M5L results do not deteriorate when increasing the dataset size, making it a candidate for supporting radiologists on large scale screenings and clinical programs.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2024

Review of low-cost self-driving laboratories in chemistry and materials science: the “frugal twin” concept

This review proposes the concept of a “frugal twin,” similar to a digital twin, but for physical experiments. Frugal twins range from simple toy examples to low-cost surrogates of high-cost research systems. For example, a color-mixing self-driving laboratory (SDL) can serve as a low-cost version of a costly multi-step chemical discovery SDL. Frugal twins already provide hands-on experience for SDLs with low costs and low risks. They can also offer as test beds for software prototyping (e.g., optimization, data infrastructure), and a low barrier to entry for democratizing SDLs. However, there is room for improvement. The true value of frugal twins can be realized in three core areas. Firstly, hardware and software modularity; secondly, purpose-built design (human-inspired vs. hardware-centric vs. human-in-the-loop); and thirdly state-of-the-art (SOTA) software (e.g., multi-fidelity optimization). We also describe the ethical benefits and risks that come with the democratization of science through frugal twins. For future work, we suggest ideas for new frugal twins, SDL educational course outcomes, and a classification scheme for autonomy levels.

36 MATERIALS SCIENCE↗
NASA NTRS2006

An Experimental Global Monitoring System for Rainfall-triggered Landslides using Satellite Remote Sensing Information

Landslides triggered by rainfall can possibly be foreseen in real time by jointly using rainfall intensity-duration thresholds and information related to land surface susceptibility. However, no system exists at either a national or a global scale to monitor or detect rainfall conditions that may trigger landslides due to the lack of extensive ground-based observing network in many parts of the world. Recent advances in satellite remote sensing technology and increasing availability of high-resolution geospatial products around the globe have provided an unprecedented opportunity for such a study. In this paper, a framework for developing an experimental real-time monitoring system to detect rainfall-triggered landslides is proposed by combining two necessary components: surface landslide susceptibility and a real-time space-based rainfall analysis system (http://trmm.gsfc.nasa.aov). First, a global landslide susceptibility map is derived from a combination of semi-static global surface characteristics (digital elevation topography, slope, soil types, soil texture, and land cover classification etc.) using a GIs weighted linear combination approach. Second, an adjusted empirical relationship between rainfall intensity-duration and landslide occurrence is used to assess landslide risks at areas with high susceptibility. A major outcome of this work is the availability of a first-time global assessment of landslide risk, which is only possible because of the utilization of global satellite remote sensing products. This experimental system can be updated continuously due to the availability of new satellite remote sensing products. This proposed system, if pursued through wide interdisciplinary efforts as recommended herein, bears the promise to grow many local landslide hazard analyses into a global decision-making support system for landslide disaster preparedness and risk mitigation activities across the world.

Hong, Yang↗
DOE OSTI2020

A data mining approach to investigate food groups related to incidence of bladder cancer in the BLadder cancer Epidemiology and Nutritional Determinants International Study

Abstract At present, analysis of diet and bladder cancer (BC) is mostly based on the intake of individual foods. The examination of food combinations provides a scope to deal with the complexity and unpredictability of the diet and aims to overcome the limitations of the study of nutrients and foods in isolation. This article aims to demonstrate the usability of supervised data mining methods to extract the food groups related to BC. In order to derive key food groups associated with BC risk, we applied the data mining technique C5.0 with 10-fold cross-validation in the BLadder cancer Epidemiology and Nutritional Determinants study, including data from eighteen case–control and one nested case–cohort study, compromising 8320 BC cases out of 31 551 participants. Dietary data, on the eleven main food groups of the Eurocode 2 Core classification codebook, and relevant non-diet data (i.e. sex, age and smoking status) were available. Primarily, five key food groups were extracted; in order of importance, beverages (non-milk); grains and grain products; vegetables and vegetable products; fats, oils and their products; meats and meat products were associated with BC risk. Since these food groups are corresponded with previously proposed BC-related dietary factors, data mining seems to be a promising technique in the field of nutritional epidemiology and deserves further examination.

Nutrition & Dietetics↗
DOE OSTI2022

Uncertainty Analysis in Multi‐Sector Systems: Considerations for Risk Analysis, Projection, and Planning for Complex Systems

Abstract Simulation models of multi‐sector systems are increasingly used to understand societal resilience to climate and economic shocks and change. However, multi‐sector systems are also subject to numerous uncertainties that prevent the direct application of simulation models for prediction and planning, particularly when extrapolating past behavior to a nonstationary future. Recent studies have developed a combination of methods to characterize, attribute, and quantify these uncertainties for both single‐ and multi‐sector systems. Here, we review challenges and complications to the idealized goal of fully quantifying all uncertainties in a multi‐sector model and their interactions with policy design as they emerge at different stages of analysis: (a) inference and model calibration; (b) projecting future outcomes; and (c) scenario discovery and identification of risk regimes. We also identify potential methods and research opportunities to help navigate the tradeoffs inherent in uncertainty analyses for complex systems. During this discussion, we provide a classification of uncertainty types and discuss model coupling frameworks to support interdisciplinary collaboration on multi‐sector dynamics (MSD) research. Finally, we conclude with recommendations for best practices to ensure that MSD research can be properly contextualized with respect to the underlying uncertainties.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2016

Estimating the Economic Potential of Offshore Wind in the United States

The potential for cost reduction and market deployment for offshore wind varies considerably within the United States. This analysis estimates the future economic viability of offshore wind at more than 7,000 sites under a variety of electric sector and cost reduction scenarios. Identifying the economic potential of offshore wind at a high geospatial resolution can capture the significant variation in local offshore resource quality, costs, and revenue potential. In estimating economic potential, this article applies a method initially developed in Brown et al. (2015) to offshore wind and estimates the sensitivity of results under a variety of most likely electric sector scenarios. For the purposes of this analysis, a theoretical framework is developed introducing a novel offshore resource classification system that is analogous to established resource classifications from the oil and gas sector. Analyzing economic potential within this framework can help establish a refined understanding across industries of the technology and site-specific risks and opportunities associated with future offshore wind development. The results of this analysis are intended to inform the development of the U.S. Department of Energy's offshore wind strategy.

17 WIND ENERGY↗
DOE OSTI2017

Aggressive Sampling for Multi-class to Binary Reduction with Applications to Text Classification

We address the problem of multi-class classification in the case where the number of classes is very large. We propose a double sampling strategy on top of a multi-class to binary reduction strategy, which transforms the original multi-class problem into a binary classification problem over pairs of examples. The aim of the sampling strategy is to overcome the curse of long-tailed class distributions exhibited in majority of large-scale multi-class classification problems and to reduce the number of pairs of examples in the expanded data. We show that this strategy does not alter the consistency of the empirical risk minimization principle defined over the double sample reduction. As a result, experiments are carried out on DMOZ and Wikipedia collections with 10,000 to 100,000 classes where we show the efficiency of the proposed approach in terms of training and prediction time, memory consumption, and predictive performance with respect to state-of-the-art approaches.

97 MATHEMATICS AND COMPUTING↗