SEARCH · Search NASA

Results for “classifiers”

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.

SourceResearch connectionFrom yearThrough year

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

At least 91 records · Page 5

All sources NASA NTRS DOE OSTI

DOE OSTI2024

Training toward significance with the decorrelated event classifier transformer neural network

Experimental particle physics uses machine learning for many tasks, where one application is to classify signal and background events. This classification can be used to bin an analysis region to enhance the expected significance for a mass resonance search. In natural language processing, one of the leading neural network architectures is the transformer. In this work, an event classifier transformer is proposed to bin an analysis region, in which the network is trained with special techniques. The techniques developed here can enhance the significance and reduce the correlation between the network’s output and the reconstructed mass. It is found that this trained network can perform better than boosted decision trees and feed-forward networks. Published by the American Physical Society 2024

Astronomy & Astrophysics↗

DOE OSTI2024

Machine Learning (ML) Classifier to Assist Metadata Creation

The Atmospheric Radiation Measurement (ARM) Data Center is responsible for the timely collection, archival, and curation of science data products. These products are freely available through an online data repository. Metadata creation is paramount for scientific users to find and access over seven petabytes of atmospheric science data. The hierarchical metadata structure allows users to search for information at both broad and narrow levels. This project aims to leverage 30 years’ worth of manually created metadata to enable machine predictions of broad-term classifications from narrow-term descriptions. These classification predictions would assist metadata coordinators with their term selections. This paper discusses the cleaning and preprocessing of the training data, the pipeline developed to determine the best model for this task, and the creation of an API metadata classifier for ARM measurement metadata. Our results show that the Linear Support Vector Classification (LinearSVC) algorithm, along with the Term Frequency – Inverse Document Frequency (TF-IDF) vectorizer, is well-suited for our multi-class classification task. Lengthier input training data led to better results, and artificial balancing was unnecessary for this particular use case. This predictive classifier enhances efficiency in metadata creation, as well as supports greater consistency and accuracy in metadata tagging.

Collier, Hannah [ORNL] (ORCID:0000000341284292)↗

DOE OSTI2026

Global Explainability of A Deep Abstaining Classifier for Cancer Pathology Reports

We present a global explainability method to characterize sources of errors in a real-world multitask deep abstaining classifier (DAC), in the context of cancer histology prediction. Our multitask classifier, currently deployed for automated annotation of cancer pathology reports from NCI-SEER registries, was trained and evaluated on 1.04 million hand-annotated samples and makes simultaneous predictions of cancer site, subsite, histology, laterality, and behavior for each report. The DAC framework enables the model to abstain on ambiguous reports and confusing classes to achieve the target accuracy on the retained (non-abstained) samples, but at the cost of decreased coverage. Requiring 97% accuracy on the histology task caused our model to retain only 22% of all samples, mostly the less ambiguous and common classes. Local explainability with the GradInp technique provided a computationally efficient way of obtaining contextual reasoning for hundreds of thousands of individual predictions. Our method, involving dimensionality reduction of approximately 13000 aggregated local explanations (ALE), offers a tractable path to true global explainability. It enabled identification of sources of errors in histology classification, globally, as hierarchical complexity among classes, label noise, insufficient information, and conflicting evidence. This suggests several strategies for iterative improvement of our DAC, including well-designed exclusion criteria, focused annotation, and reduced penalties for errors involving hierarchically related classes.

59 BASIC BIOLOGICAL SCIENCES↗

DOE OSTI2022

Meta-optic accelerators for object classifiers

Rapid advances in deep learning have led to paradigm shifts in a number of fields, from medical image analysis to autonomous systems. These advances, however, have resulted in digital neural networks with large computational requirements, resulting in high energy consumption and limitations in real-time decision-making when computation resources are limited. Here, we demonstrate a meta-optic–based neural network accelerator that can off-load computationally expensive convolution operations into high-speed and low-power optics. In this architecture, metasurfaces enable both spatial multiplexing and additional information channels, such as polarization, in object classification. End-to-end design is used to co-optimize the optical and digital systems, resulting in a robust classifier that achieves 93.1% accurate classification of handwriting digits and 93.8% accuracy in classifying both the digit and its polarization state. This approach could enable compact, high-speed, and low-power image and information processing systems for a wide range of applications in machine vision and artificial intelligence.

42 ENGINEERING↗

DOE OSTI2021

Classified library celebrates two-year anniversary print your own commemorative history posters

Today is the National Security Research Center’s second anniversary. These two years, though, have been eight decades in the making, said NSRC Director Riz Ali. “We opened our doors in 2019, but the Center’s beginnings date back to World War II when J. Robert Oppenheimer created the Lab in Los Alamos to develop the first-ever nuclear weapons,” Ali said. “An important part of opening the Lab included standing up a technical reports library. Since then, we have amassed millions of holdings and today we are one of the top research institutions in the country.” What had long-been a repository of classified archival materials is today a dynamic, vibrant library that outsizes the collections of other similar research libraries around the country. Today, the NSRC houses the Laboratory’s classified nuclear weapons research, data and reports, as well as unclassified historical artifacts – the stories, photos, films and more from the people who have been a part of the Lab from the start.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗

DOE OSTI2020

Proteomic Tissue-Based Classifier for Early Prediction of Prostate Cancer Progression

Although ~40% of screen-detected prostate cancers (PCa) are indolent, advanced-stage PCa is a lethal disease with 5-year survival rates around 29%. Identification of biomarkers for early detection of aggressive disease is a key challenge. Starting with 52 candidate biomarkers, selected from existing PCa genomics datasets and known PCa driver genes, we used targeted mass spectrometry to quantify proteins that significantly differed in primary tumors from PCa patients treated with radical prostatectomy (RP) across three study outcomes: (i) metastasis ≥1-year post-RP, (ii) biochemical recurrence ≥1-year post-RP, and (iii) no progression after ≥10 years post-RP. Sixteen proteins that differed significantly in an initial set of 105 samples were evaluated in the entire cohort (n = 338). A five-protein classifier which combined FOLH1, KLK3, TGFB1, SPARC, and CAMKK2 with existing clinical and pathological standard of care variables demonstrated significant improvement in predicting distant metastasis, achieving an area under the receiver-operating characteristic curve of 0.92 (0.86, 0.99, p = 0.001) and a negative predictive value of 92% in the training/testing analysis. This classifier has the potential to stratify patients based on risk of aggressive, metastatic PCa that will require early intervention compared to low risk patients who could be managed through active surveillance.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2021

Analysis of Previously Classified White Dwarf–Main-sequence Binaries Using Data from the APOGEE Survey

We present analyses of near-infrared spectroscopic data from the Apache Point Observatory Galactic Evolution Experiment (APOGEE) survey for 45 previously confirmed or candidate white dwarf–main-sequence (WDMS) binaries identified by the optical Sloan Digital Sky Survey (SDSS) and LAMOST surveys. Among these 45 systems, we classify three as having red giant primaries in the LAMOST sample and 14 as young stellar object contaminants in the photometrically identified SDSS sample. From among the subsample of 28 systems that we confirm to have MS primaries, we derive and place limits on orbital periods and velocity amplitudes for 14. Seven systems have significant velocity variations that warrant a post-common-envelope (PCE) binary classification, four of which are newly classified, three of which are newly confirmed, and five for which we can derive full orbital parameters. If confirmed, one of these newly discovered systems (2M14544500+4626456) will have the second-longest orbital period reported for a typical compact PCE WDMS binary (P = 15.1 days). In addition to the seven above, we also recover and characterize with APOGEE data the well-known PCE WDMS systems EG UMa and HZ 9. We also investigate the overall metallicity distribution of the WDMS sample, which is a parameter space not often explored for these systems. Of note, we find one system (2M14244053+4929580) to be extremely metal-poor ([Fe/H] = − 1.42) relative to the rest of the near-solar sample. Additionally, the PCE systems in our sample are found to be, on average, higher in metallicity than their wide-binary counterparts, though we caution that with this small number of systems, the sample may not be representative of the overall distribution of WDMS systems.

36 MATERIALS SCIENCE↗

DOE OSTI2025

Classifying thermodynamic cloud phase using machine learning models

Vertically resolved thermodynamic cloud-phase classifications are essential for studies of atmospheric cloud and precipitation processes. The Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) Thermodynamic Cloud Phase (THERMOCLDPHASE) value-added product (VAP) uses a multi-sensor approach to classify the thermodynamic cloud phase by combining lidar backscatter and depolarization, radar reflectivity, Doppler velocity, spectral width, microwave-radiometer-derived liquid water path, and radiosonde temperature measurements. The measured pixels are classified as ice, snow, mixed phase, liquid (cloud water), drizzle, rain, and liq_driz (liquid+drizzle). We use this product as the ground truth to train three machine learning (ML) models to predict the thermodynamic cloud phase from multi-sensor remote sensing measurements taken at the ARM North Slope of Alaska (NSA) observatory: a random forest (RF), a multi-layer perceptron (MLP), and a convolutional neural network (CNN) with a U-Net architecture. Evaluations against the outputs of the THERMOCLDPHASE VAP with 1 year of data show that the CNN outperforms the other two models, achieving the highest test accuracy, F1 score, and mean intersection over union (IOU). Analysis of ML confidence scores shows that ice, rain, and snow have higher confidence scores, followed by liquid, while mixed, drizzle, and liq_driz have lower scores. Feature importance analysis reveals that the mean Doppler velocity and vertically resolved temperature are the most influential data streams for ML thermodynamic cloud-phase predictions. Lidar measurements exhibit lower feature importance due to rapid signal attenuation caused by the frequent presence of persistent low-level clouds at the NSA site. The ML models' generalization capacity is further evaluated by applying them at another Arctic ARM site in Norway using data taken during the ARM Cold-Air Outbreaks in the Marine Boundary Layer Experiment (COMBLE) field campaign. The models demonstrated similar performance to that observed at the NSA site. Finally, we evaluate the ML models' response to simulated instrument outages and signal degradation and show that a CNN U-Net model trained with input channel dropouts performs better when input fields are missing.

ARM Aerial Facility↗

NASA NTRS1970

Stellar spectrum classifier

Stellar classification is accomplished by correlating holograms of classified spectra with the Fourier transform of the spectral images to be classified.

Reid, J. H.↗

NASA NTRS1971

Layered 'recognition cone' networks that pre-process, classify, and describe.

A sequence of six types of pattern recognition system is examined. A program is described to illustrate some of the features developed. The first type (similar to many of the programs currently used) preprocesses by applying layers of local averaging and differencing transforms to smooth, fill in gaps and heighten contours, curves, and angles. It then applies a set of characterizers, each of which implies a set of names. The program chooses the single most high implied name. The second type combines the preprocessing transforms and the characterizers into a single operation of general type. Transforms build up a next representation of the input, while the characterizers imply the output name. The third type erases the distinction between a transform and an implication. Now all outputs are stored in the next transform layer. As the program averages information, its layers shrink, so that the system builds a cone of layers. When the program reaches the apex (a layer of only one cell that contains all the information), it chooses the single name with which it classifies the input. The fourth type is capable of choosing more than one name and, therefore, can both describe and classify the scene. The fifth type examines the interrelations among the set of names chosen. The sixth step can be taken to converse about the scene, developing an appropriate description in response to suggestions and queries. This allows the program to perform more computations and to look again on demand.

Uhr, L.↗

NASA NTRS1975

A hybrid classifier using the parallelepiped and Bayesian techniques

A versatile classification scheme is developed which uses the best features of the parallelepiped algorithm and the Bayesian maximum likelihood algorithm. The parallelepiped technique has the advantage of being very fast, especially when implemented into a table look-up scheme; its disadvantage is its inability to distinguish and classify spectral signatures which are similar in nature. This disadvantage is eliminated by the Bayesian technique which is capable of distinguishing subtle differences very well. The hybrid algorithm developed reduces computer time by as much as 90%. A two- and n-dimensional description of the hybrid classifier is given.

Addington, J. D.↗

NASA NTRS1975

Adaptive Bayes classifiers for remotely sensed data

An algorithm is developed for a learning, adaptive, statistical pattern classifier for remotely sensed data. The estimation procedure consists of two steps: (1) an optimal stochastic approximation of the parameters of interest, and (2) a projection of the parameters in time and space. The results reported are for Gaussian data in which the mean vector of each class may vary with time or position after the classifier is trained.

Raulston, H. S.↗

NASA NTRS1978

A Determination of the Optimum Time of Year for Remotely Classifying Marsh Vegetation from LANDSAT Multispectral Scanner Data

The author has identified the following significant results. A technique was used to determine the optimum time for classifying marsh vegetation from computer-processed LANDSAT MSS data. The technique depended on the analysis of data derived from supervised pattern recognition by maximum likelihood theory. A dispersion index, created by the ratio of separability among the class spectral means to variability within the classes, defined the optimum classification time. Data compared from seven LANDSAT passes acquired over the same area of Louisiana marsh indicated that June and September were optimum marsh mapping times to collectively classify Baccharis halimifolia, Spartina patens, Spartina alterniflora, Juncus roemericanus, and Distichlis spicata. The same technique was used to determine the optimum classification time for individual species. April appeared to be the best month to map Juncus roemericanus; May, Spartina alterniflora; June, Baccharis halimifolia; and September, Spartina patens and Distichlis spicata. This information is important, for instance, when a single species is recognized to indicate a particular environmental condition.

Butera, M. K.↗

NASA NTRS1979

Fisher classifier and its probability of error estimation

Computationally efficient expressions are derived for estimating the probability of error using the leave-one-out method. The optimal threshold for the classification of patterns projected onto Fisher's direction is derived. A simple generalization of the Fisher classifier to multiple classes is presented. Computational expressions are developed for estimating the probability of error of the multiclass Fisher classifier.

Chittineni, C. B.↗

NASA NTRS1979

Software requirements for the study of contextual classifiers and label imperfections

The software requirements for the study of contextual classifiers and imperfections in the labels are presented. In particular, the requirements are described for updating the posteriori probability of the picture element under consideration using information from its local neighborhood, designing the Fisher classifier, and other required routines. Only the necessary equations are given for the development of software.

Chittineni, C. B.↗

NASA NTRS1979

Development of a tree classifier for discrimination of surface mine activity from Landsat digital data

In a cooperative project with the Geological Survey of Alabama, the Mississippi State Remote Sensing Applications Program has developed a single purpose, decision-tree classifier using band-ratioing techniques to discriminate various stages of surface mining activity. The tree classifier has four levels and employs only two channels in classification at each level. An accurate computation of the amount of disturbed land resulting from the mining activity can be made as a product of the classification output. The utilization of Landsat data provides a cost-efficient, rapid, and accurate means of monitoring surface mining activities.

Solomon, J. L.↗

NASA NTRS1980

Parallel processing implementations of a contextual classifier for multispectral remote sensing data

The applicability of parallel processing schemes to the implementation of a contextual classification algorithm which exploits the spatial and spectral context of a multispectral remote sensing pixel to achieve classification is examined. Two algorithms for classifying each multivariate pixel taking into account the probable classifications of neighboring pixels are presented which make use of a size three horizontally linear neighborhood, and the serial computational complexity of the more efficient algorithm is shown to grow in proportion to the number of pixels and the cube of the number of possible categories. The implementation of the more efficient algorithm on a CDC Flexible Processor system and on a multimicroprocessor system such as the proposed PASM is then discussed. It is noted that the use of N processors to perform the calculations N times faster than a single processor overcomes the principal disadvantage of contexual classifiers, i.e., their computational complexity.

Siegel, H. J.↗

NASA NTRS1986

On the design of classifiers for crop inventories

Crop proportion estimators that use classifications of satellite data to correct, in an additive way, a given estimate acquired from ground observations are discussed. A linear version of these estimators is optimal, in terms of minimum variance, when the regression of the ground observations onto the satellite observations in linear. When this regression is not linear, but the reverse regression (satellite observations onto ground observations) is linear, the estimator is suboptimal but still has certain appealing variance properties. In this paper expressions are derived for those regressions which relate the intercepts and slopes to conditional classification probabilities. These expressions are then used to discuss the question of classifier designs that can lead to low-variance crop proportion estimates. Variance expressions for these estimates in terms of classifier omission and commission errors are also derived.

Heydorn, R. P.↗