Search NASA⌕ Search

SEARCH · Search NASA

Results for “word vectors”

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 19 records
All sourcesNASA NTRSDOE OSTI
NASA NTRSDate not supplied

ES2Vec: Earth Science Metadata Suggestions and Analogical Reasoning

As the volume of text-based Earth science research grows, it is increasingly possible to discover latent relationships in the literature. However, traditional methodologies are restricted by limited computational capabilities and intractable problem spaces. Advancements in natural language processing (NLP) have allowed us to use an extensive Earth science corpus to create a domain-specific word vector model, Es2Vec, which we have used to surface latent relationships between Earth science concepts and generate improved keyword tags. Earth science metadata keyword assignment is a challenging problem. Dataset curators select appropriate keywords from the Global Change Master Directory (GCMD) set of keywords. The keywords an are integral part of the search and discovery of these datasets. Hence, the selection of keywords is crucial to increasing the discoverability of datasets. Utilizing machine learning techniques, we provide users with automated keyword suggestions to complement manual selection. We trained a machine learning model that leverages the semantic embedding ability of Word2Vec models to process abstracts and suggest relevant keywords. A user interface tool we built to assist data curators in the assignment of such keywords is also described.

word vectors↗
NASA NTRS2019

Predictive Modeling for Differential Diagnosis and Mortality Risk Assessment

The prevalence of electronic health record (EHR) systems has brought prodigious biomedical informatics opportunity. Automated machine learning methods can effectively utilize such data and have become common tools for healthcare predictive modeling. Researches in medical informatics have explored the potential of deep learning and classical models in emergent care scenarios. In particular, predicting differential diagnoses for admissions have proven useful in decreasing unnecessary lab tests and improving inpatient triage decision-making. Moreover, identification of high-risk patients for in-hospital mortality is vitally important to maximize allocation of medical resources.The Medical Information Mart for Intensive Care (MIMIC-III) database, containing de-identified critical care inpatient was used in our study. This data set captures hospital patient laboratory measurements, pharmacologic prescriptions, diagnostic data and procedure event recordings. When considering adult patients and discounting admissions with ICU length of stay less than 24 hours, there were 37,787 unique admissions and 30,414 total patients. We examined the top 25 most prevalent ICD-9 group-level disease specificities in MIMIC-III using a multi-label classification model. In-hospital mortality was modeled as binary classification with 4,155 (13%) adult patients that expired, of which 3,138 (75.5%) were in the ICU setting. The metrics AUC, F1 score, sensitivity and specificity values calculated for each disease label measured prediction performance.The usage of ICD-9 group codes reduced feature dimension from 14,567 to 942 and greatly improved distribution of patient diagnostic categories. Disease temporal patterns were captured by considering the most frequently sampled 6 vital signs and 13 laboratory values. Missing data were imputed at each time-stamp. Time-series raw hourly average values were converted into 5 summary features (mean, standard deviation, number of observations, min & max values). Patient demographic variables such as age, gender, marital status and ethnicity were also factored into the modeling. Choi et al showed that contextual embedding of medical data, diagnostic and procedural codes alone can predict future diagnoses with sensitivity as high as 0.79. We utilized an embedding technique called word2vec which allowed sparse representations of medical history to be transformed into dense word vectors. The mappings captured contextual information by treating each admission as a sentence and learning the most likely neighboring words in a sliding window fashion. Binary and multi-label classification was achieved via collapse models, which do not consider temporal information, as well as recurrent neural networks with regularization, Softmax output layer activation together with categorical cross-entropy as the loss function.

US Army collaboration↗
NASA NTRS2009

Bounded-Angle Iterative Decoding of LDPC Codes

Bounded-angle iterative decoding is a modified version of conventional iterative decoding, conceived as a means of reducing undetected-error rates for short low-density parity-check (LDPC) codes. For a given code, bounded-angle iterative decoding can be implemented by means of a simple modification of the decoder algorithm, without redesigning the code. Bounded-angle iterative decoding is based on a representation of received words and code words as vectors in an n-dimensional Euclidean space (where n is an integer).

Dolinar, Samuel↗
NASA NTRS1995

VLSI Processor For Vector Quantization

Pixel intensities in each kernel compared simultaneously with all code vectors. Prototype high-performance, low-power, very-large-scale integrated (VLSI) circuit designed to perform compression of image data by vector-quantization method. Contains relatively simple analog computational cells operating on direct or buffered outputs of photodetectors grouped into blocks in imaging array, yielding vector-quantization code word for each such block in sequence. Scheme exploits parallel-processing nature of vector-quantization architecture, with consequent increase in speed.

Tawel, Raoul↗
NASA NTRS1988

Robust vector quantization for noisy channels

The paper briefly discusses techniques for making vector quantizers more tolerant to tranmsission errors. Two algorithms are presented for obtaining an efficient binary word assignment to the vector quantizer codewords without increasing the transmission rate. It is shown that about 4.5 dB gain over random assignment can be achieved with these algorithms. It is also proposed to reduce the effects of error propagation in vector-predictive quantizers by appropriately constraining the response of the predictive loop. The constrained system is shown to have about 4 dB of SNR gain over an unconstrained system in a noisy channel, with a small loss of clean-channel performance.

Demarca, J. R. B.↗
NASA NTRS1982

Evaluation of the SPAR thermal analyzer on the CYBER-203 computer

The use of the CYBER 203 vector computer for thermal analysis is investigated. Strengths of the CYBER 203 include the ability to perform, in vector mode using a 64 bit word, 50 million floating point operations per second (MFLOPS) for addition and subtraction, 25 MFLOPS for multiplication and 12.5 MFLOPS for division. The speed of scalar operation is comparable to that of a CDC 7600 and is some 2 to 3 times faster than Langley's CYBER 175s. The CYBER 203 has 1,048,576 64-bit words of real memory with an 80 nanosecond (nsec) access time. Memory is bit addressable and provides single error correction, double error detection (SECDED) capability. The virtual memory capability handles data in either 512 or 65,536 word pages. The machine has 256 registers with a 40 nsec access time. The weaknesses of the CYBER 203 include the amount of vector operation overhead and some data storage limitations. In vector operations there is a considerable amount of time before a single result is produced so that vector calculation speed is slower than scalar operation for short vectors.

Robinson, J. C.↗
NASA NTRS1991

Analysis of structural response data using discrete modal filters

The application of reciprocal modal vectors to the analysis of structural response data is described. Reciprocal modal vectors are constructed using an existing experimental modal model and an existing frequency response matrix of a structure, and can be assembled into a matrix that effectively transforms the data from the physical space to a modal space within a particular frequency range. In other words, the weighting matrix necessary for modal vector orthogonality (typically the mass matrix) is contained within the reciprocal model matrix. The underlying goal of this work is mostly directed toward observing the modal state responses in the presence of unknown, possibly closed loop forcing functions, thus having an impact on both operating data analysis techniques and independent modal space control techniques. This study investigates the behavior of reciprocol modal vectors as modal filters with respect to certain calculation parameters and their performance with perturbed system frequency response data.

Freudinger, Lawrence C.↗
NASA NTRS1990

Reliable and Efficient Parallel Processing Algorithms and Architectures for Modern Signal Processing

Least-squares (LS) estimations and spectral decomposition algorithms constitute the heart of modern signal processing and communication problems. Implementations of recursive LS and spectral decomposition algorithms onto parallel processing architectures such as systolic arrays with efficient fault-tolerant schemes are the major concerns of this dissertation. There are four major results in this dissertation. First, we propose the systolic block Householder transformation with application to the recursive least-squares minimization. It is successfully implemented on a systolic array with a two-level pipelined implementation at the vector level as well as at the word level. Second, a real-time algorithm-based concurrent error detection scheme based on the residual method is proposed for the QRD RLS systolic array. The fault diagnosis, order degraded reconfiguration, and performance analysis are also considered. Third, the dynamic range, stability, error detection capability under finite-precision implementation, order degraded performance, and residual estimation under faulty situations for the QRD RLS systolic array are studied in details. Finally, we propose the use of multi-phase systolic algorithms for spectral decomposition based on the QR algorithm. Two systolic architectures, one based on triangular array and another based on rectangular array, are presented for the multiphase operations with fault-tolerant considerations. Eigenvectors and singular vectors can be easily obtained by using the multi-pase operations. Performance issues are also considered.

Liu, Kuojuey Ray↗
NASA NTRS1984

Navier-Stokes Simulation of Homogeneous Turbulence on the CYBER 205

A computer code which solves the Navier-Stokes equations for three dimensional, time-dependent, homogenous turbulence has been written for the CYBER 205. The code has options for both 64-bit and 32-bit arithmetic. With 32-bit computation, mesh sizes up to 64 (3) are contained within core of a 2 million 64-bit word memory. Computer speed timing runs were made for various vector lengths up to 6144. With this code, speeds a little over 100 Mflops have been achieved on a 2-pipe CYBER 205. Several problems encountered in the coding are discussed.

Wu, C. T.↗
NASA NTRS1988

Navier-Stokes calculations for DFVLR F5-wing in wind tunnel using Runge-Kutta time-stepping scheme

A three-dimensional Navier-Stokes code using an explicit multistage Runge-Kutta type of time-stepping scheme is used for solving the transonic flow past a finite wing mounted inside a wind tunnel. Flow past the same wing in free air was also computed to assess the effect of wind-tunnel walls on such flows. Numerical efficiency is enhanced through vectorization of the computer code. A Cyber 205 computer with 32 million words of internal memory was used for these computations.

Vatsa, V. N.↗
NASA NTRS1992

SIMBAD quality-control

The astronomical database SIMBAD developed at the Centre de donnees astronomiques de Strasbourg presently contains 760,000 objects (stellar and non-stellar). It has the unique characteristic of being structured specifically for astronomical objects. All types of heterogeneous data (bibliographic references, measurements, and sets of identification) are connected with each object. The attributes that define quality of the database include the following. Reliability: cross-identification should not rely upon just exact values object coordinates. It also means that information attached to one simple object should be consistent. The existing data must be controlled in order to start with a reliable base and to cross-identify new data assuring the quality as data grows. Exhaustivity: delays between publication of new informations and their inclusion in the database should be as short as possible. The integrity of the database has to be maintained as data accumulates. Taking the amount of data into consideration and the rate of new data production, it is necessary to use automatic methods. One of the possibilities is to use multivariate data analysis. The factor-space is a n-dimensional relevancy space which is described by the n-axes representing a set of n subject matter headings; the words and phrases can be used to scale the axes and the documents are then a vector average of the terms within them. The application reported herein is based on the NASA-STI bibliographical database. The selected data concern astronomy, astrophysics, and space radiation (102,963 references from 1975 to 1991 included 8070 keywords). The F-space is built from this bibliographical data. By comparing the F-space position obtained from the NASA-STI keywords with the F-space position obtained from the SIMBAD references, the authors will be able to show whether it is possible to retrieve information with a restricted set of words only. If the comparison is valid, this will be a way to enter bibliographic information in the SIMBAD quality control process. Furthermore, it is possible to connect the physical measurements of stars from SIMBAD to literature concerning these stars from the NASA-STI abstracts. The physical properties of stars (e.g. UBV colors) are not randomly distributed. Stars are distributed among different clusters in a physical parameter space. The authors will show that there are some relations between this classification and the literature concerning these objects clusters in a factor space. They will investigate the nature of the relationship between the SIMBAD measurements and the bibliography. These would be new relationships that are not pre-established by an astronomer. In addition, the bibliography could be neutral information that can be used in combination with the measured parameters.

Lesteven, Soizick↗
NASA NTRS1997

Head-Mounted and Head-Up Display Glossary

One of the problems in head-up and helmet-mounted display (HMD) literature has been a lack of standardization of words and abbreviations. Several different words have been used for the same concept; for example, flight path angle, flight path marker, velocity vector, and total velocity vector all refer to the same thing. In other cases, the same term has been used with two different meanings, such as binocular field-of-view which means the field-of-view visible to both left and right eyes according to some or the field-of-view visible to either the left or right eye or both according to others. Many of the terms used in HMD studies have not been well-defined. We need to have a common language to ensure that system descriptions are communicated. As an example, the term 'stabilized' has been widely used with two meanings. 'Roll-stabilized' has been used to mean a symbol which rotates to indicate the roll or bank of the aircraft. 'World-stabilized' and 'head-stabilized' have both been used to indicate symbols which move to remain fixed with respect to external objects. HMDs present unique symbology problems not found in HUDs. Foremost among these is the issue of maintaining spatial orientation of the symbols. All previous flight displays, round dial instruments, HDDs, and HUDs have been fixed in the cockpit. With the HMD, the flight display can move through a large angle. The coordinates use in transforming from the real-world to the aircraft to the HMD have not been consistently defined. This glossary contains terms relating to optics and vision, displays, and flight information, weapons and aircraft systems. Some definitions, such as Navigation Display, have been added to clarify the definitions for Primary Flight Display and Primary Flight Reference. A list of HUD/HMD related abbreviations is also included.

Newman, Richard L.↗
NASA NTRS1985

Numerical solutions of 3-dimensional Navier-Stokes equations for closed bluff-bodies

The Navier-Stokes equations are solved numerically. These equations are unsteady, compressible, viscous, and three-dimensional without neglecting any terms. The time dependency of the governing equations allows the solution to progress naturally for an arbitrary initial guess to an asymptotic steady state, if one exists. The equations are transformed from physical coordinates to the computational coordinates, allowing the solution of the governing equations in a rectangular parallelepiped domain. The equations are solved by the MacCormack time-split technique which is vectorized and programmed to run on the CDc VPS 32 computer. The codes are written in 32-bit (half word) FORTRAN, which provides an approximate factor of two decreasing in computational time and doubles the memory size compared to the 54-bit word size.

Abolhassani, J. S.↗
NASA NTRS2020

Amino Acid Encoding for Deep Learning Applications

Background: The number of applications of deep learning algorithms in bioinformatics is increasing as they usually achieve superior performance over classical approaches, especially, when bigger training datasets are available. In deep learning applications, discrete data, e.g. words or n-grams in language, or amino acids or nucleotides in bioinformatics, are generally represented as a continuous vector through an embedding matrix. Recently, learning this embedding matrix directly from the data as part of the continuous iteration of the model to optimize the target prediction – a process called ‘end-to-end learning’ – has led to state-of-the-art results in many fields. Although usage of embeddings is well described in the bioinformatics literature, the potential of end-to-end learning for single amino acids, as compared to more classical manually-curated encoding strategies, has not been systematically addressed. To this end, we compared classical encoding matrices, namely one-hot, VHSE8 and BLOSUM62, to end-to-end learning of amino acid embeddings for two different prediction tasks using three widely used architectures, namely recurrent neural networks (RNN), convolutional neural networks (CNN), and the hybrid CNN-RNN. Results: By using different deep learning architectures, we show that end-to-end learning is on par with classical encodings for embeddings of the same dimension even when limited training data is available, and might allow for a reduction in the embedding dimension without performance loss, which is critical when deploying the models to devices with limited computational capacities. We found that the embedding dimension is a major factor in controlling the model performance. Surprisingly, we observed that deep learning models are capable of learning from random vectors of appropriate dimension. Conclusion: Our study shows that end-to-end learning is a flexible and powerful method for amino acid encoding. Further, due to the flexibility of deep learning systems, amino acid encoding schemes should be benchmarked against random vectors of the same dimension to disentangle the information content provided by the encoding scheme from the distinguishability effect provided by the scheme.

Deep-learning↗
NASA NTRS2020

Amino Acid Encoding for Deep Learning Applications

Background: The number of applications of deep learning algorithms in bioinformatics is increasing as they usually achieve superior performance over classical approaches, especially, when bigger training datasets are available. In deep learning applications, discrete data, e.g. words or n-grams in language, or amino acids or nucleotides in bioinformatics, are generally represented as a continuous vector through an embedding matrix. Recently, learning this embedding matrix directly from the data as part of the continuous iteration of the model to optimize the target prediction – a process called ‘end-to-end learning’ – has led to state-ofthe-art results in many fields. Although usage of embeddings is well described in the bioinformatics literature, the potential of end-to-end learning for single amino acids, as compared to more classical manually-curated encoding strategies, has not been systematically addressed. To this end, we compared classical encoding matrices, namely one-hot, VHSE8 and BLOSUM62, to end-to-end learning of amino acid embeddings for two different prediction tasks using three widely used architectures, namely recurrent neural networks (RNN), convolutional neural networks (CNN), and the hybrid CNN-RNN. Results: By using different deep learning architectures, we show that end-to-end learning is on par with classical encodings for embeddings of the same dimension even when limited training data is available, and might allow for a reduction in the embedding dimension without performance loss, which is critical when deploying the models to devices with limited computational capacities. We found that the embedding dimension is a major factor in controlling the model performance. Surprisingly, we observed that deep learning models are capable of learning from random vectors of appropriate dimension. Conclusion: Our study shows that end-to-end learning is a flexible and powerful method for amino acid encoding. Further, due to the flexibility of deep learning systems, amino acid encoding schemes should be benchmarked against random vectors of the same dimension to disentangle the information content provided by the encoding scheme from the distinguishability effect provided by the scheme.

Hesham ElAbd↗
NASA NTRS1981

Singular perturbation, state aggregation and nonlinear filtering

Consideration is given to a state process evolving in R(n), whose motion is that of a pure jump process in R(n) in the 0(1) time scale, upon which is superimposed a continuous motion along the orbits of a gradient-like vector field g in R(n) in the 0(1/epsilon) time scale. The infinitesimal generator of the state process is, in other words, of the form L + (1/epsilon)g. It follows from the main results presented that the projected filters converge to the finite state Wonham filter corresponding to the problem of estimating the finite state process in the presence of additive white noise.

Hijab, O.↗
NASA NTRS1983

Three-dimensional flow over a conical afterbody containing a centered propulsive jet - A numerical simulation

The supersonic flow field over a body of revolution incident to the free stream is simulated numerically on a large, array processor (the CDC Cyber 205). The configuration is composed of a cone-cylinder forebody followed by a conical afterbody from which emanates a centered, supersonic propulsive jet. The free-stream Mach number is 2, the jet-exit Mach number is 2.5, and the jet-to-free-stream static pressure ratio is 3. Both the external flow and the exhaust are ideal air at a common total temperature. The thin-layer approximation to the time-dependent, compressible, Reynolds-averaged Navier-Stokes equations are solved using an implicit finite-difference algorithm. The data base, of 5 million words, is structured in a 'pencil' format so that efficient use of the array processor can be realized. The computer code is completely vectorized to take advantage of the data structure. Turbulence closure is achieved using an empirical algebraic eddy-viscosity model. The configuration and flow conditions correspond to published experimental tests and the computed solutions are consistent with the experimental data.

Deiwert, G. S.↗