Search NASA⌕ Search

SEARCH · Search NASA

Results for “unsupervised method”

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 199 records · Page 11

Optimizing training trajectories in variational autoencoders via latent Bayesian optimization approach *

Unsupervised and semi-supervised ML methods such as variational autoencoders (VAE) have become widely adopted across multiple areas of physics, chemistry, and materials sciences due to their capability in disentangling representations and ability to find latent manifolds for classification and/or regression of complex experimental data. Like other ML problems, VAEs require hyperparameter tuning, e.g. balancing the Kullback–Leibler and reconstruction terms. However, the training process and resulting manifold topology and connectivity depend not only on hyperparameters, but also their evolution during training. Because of the inefficiency of exhaustive search in a high-dimensional hyperparameter space for the expensive-to-train models, here we have explored a latent Bayesian optimization (zBO) approach for the hyperparameter trajectory optimization for the unsupervised and semi-supervised ML and demonstrated for joint-VAE with rotational invariances. We have demonstrated an application of this method for finding joint discrete and continuous rotationally invariant representations for modified national institute of standards and technology database (MNIST) and experimental data of a plasmonic nanoparticles material system. The performance of the proposed approach has been discussed extensively, where it allows for any high dimensional hyperparameter trajectory optimization of other ML models.

42 ENGINEERING↗

Unsupervised Learning Based Interaction Force Model for Nonspherical Particles in Incompressible Flows

This project provides a neural network-based interaction force model for gas-solid flows from low to intermediate Reynolds numbers and concentration, which can be linked to MFiX-DEM. We have constructed a database of the interaction force between the irregular-shaped particles using a spherical harmonic method and the fluid phase based on the particle-resolved direct numerical simulation (PR-DNS) with immersed boundary-based gas kinetic scheme. Unsupervised learning method, i.e., variational auto-encoder (VAE) has been applied to extract the primitive shape factors determining the drag force, lifting forces, and torque. The interaction force model has been trained and validated with a simple but effective multi-layer feed-forward neural network: multi-layer perceptron (MLP), which will be concatenated after the encoder of the previously trained VAE for geometry feature extraction for single, irregular particles. We have trained transpose convolutional neural networks with the PR-DNS data to predict the velocity and pressure gradient of the single particle systems and utilized them to calculate drag force of multi-particle systems. This model can provide high computational efficiency because it does not require collecting multiparticle system data from PR-DNS.

99 GENERAL AND MISCELLANEOUS↗

Spread Spectrum Time Domain Reflectometry (SSTDR) and Frequency Domain Reflectometry (FDR) for Detection of Cable Anomalies Using Machine Learning

Cables are initially qualified for nuclear power plant use for 40 years. As plants extend their operating license to 60 and 80 years, continued use of these cables must shift to a performance-based approach since it is cost prohibitive to completely replace cables that are likely still capable of performing their design function. A variety of cable tests are available and are commonly applied during outages when the cables can be taken out of service. Frequency domain reflectometry (FDR) is one of these test methods that is being more broadly accepted and used because it not only detects anomalies along the cable with a low-voltage signal that does not stress the cable insulation, but the technique also locates the anomalies. This supports follow-up local inspection and local repair or partial replacement of a damaged cable segment. Currently, FDR testing is only applied to cables that are taken out of service since the test instrument would be damaged by operational voltages. A related technology that has found some acceptance in the aircraft and rail industry is spread spectrum time domain reflectometry (SSTDR). This technology has been implemented with a custom commercial instrument by LiveWire Innovation Inc. that is designed to operate on live cables up to 1000 volts. One of the main conclusions of a previous effort was that cable reflectometry plots can be difficult for humans to analyze due to baseline noise, low or noisy anomaly response peaks, or large responses from cable ends. Detection of cable anomalies for many of these frequencies and test conditions was challenging for manual analysis. This presented an ideal opportunity for ML analysis to distinguish undamaged cable indications from anomalous cable indications. This research discusses application of machine learning (ML) to reflectometry cable test methods. The goal was to assess feasibility to distinguish undamaged cable reflectometry responses from damaged or anomalous cable reflectometry responses. The assessment considered the 3 instruments, multiple frequency bandwidths from each instrument, multiple cable anomalies and test conditions, and both supervised and unsupervised ML approaches. Although approaches and analysis methods were not identical or directly comparable, both outputs were encouraging. The unsupervised prediction weighted accuracy was assessed by instrument and by frequency. It performed better at high frequencies with the highest prediction accuracy of 0.84 for the higher frequency FDR, 0.79 for the 48-MHz LiveWire SSTDR, and 0.77 for 300-MHz PNNL SSTDR. The initial weighted accuracy average across all frequencies for using supervised ML was 0.56 to 0.68. The supervised analysis was repeated with noisier training data removed resulting in weighted accuracies of 0.69 to 0.87. These weighted accuracies are not directly comparable due to differences in the supervised and unsupervised analysis details but do indicate an encouraging trend. Even with limited and unbalanced data, strong prediction accuracies seem encouraging for further work including more data under a wider range of conditions.

42 ENGINEERING↗

Evaluating lightweight unsupervised online IDS for masquerade attacks in CAN

Vehicular controller area networks (CANs) are susceptible to masquerade attacks by malicious adversaries. In masquerade attacks, adversaries silence a targeted ID and then send malicious frames with forged content at the expected timing of benign frames. As masquerade attacks could seriously harm vehicle functionality and are the stealthiest attacks to detect in CAN, recent work has devoted attention to compare frameworks for detecting masquerade attacks in CAN. However, most existing works report offline evaluations using CAN logs already collected using simulations that do not comply with the domain’s real-time constraints. Here we contribute to advance the state of the art by presenting a comparative evaluation of four different non-deep learning (DL)-based unsupervised online intrusion detection systems (IDS) for masquerade attacks in CAN. Our approach differs from existing comparative evaluations in that we analyze the effect of controlling streaming data conditions in a sliding window setting. In doing so, we use realistic masquerade attacks being replayed from the ROAD dataset. We show that although evaluated IDS are not effective at detecting every attack type, the method that relies on detecting changes in the hierarchical structure of clusters of time series produces the best results at the expense of higher computational overhead. We discuss limitations, open challenges, and how the evaluated methods can be used for practical unsupervised online CAN IDS for masquerade attacks.

Anomaly detection↗

Combustion machine learning: Principles, progress and prospects

Progress in combustion science and engineering has led to the generation of large amounts of data from large-scale simulations, high-resolution experiments, and sensors. This corpus of data offers enormous opportunities for extracting new knowledge and insights—if harnessed effectively. Machine learning (ML) techniques have demonstrated remarkable success in data analytics, thus offering a new paradigm for data-intense analyses and scientific investigations through combustion machine learning (CombML). While data-driven methods are utilized in various combustion areas, recent advances in algorithmic developments, the accessibility of open-source software libraries, the availability of computational resources, and the abundance of data have together rendered ML techniques ubiquitous in scientific analysis and engineering. This article examines ML techniques for applications in combustion science and engineering. Starting with a review of sources of data, data-driven techniques, and concepts, we examine supervised, unsupervised, and semi-supervised ML methods. Various combustion examples are considered to illustrate and to evaluate these methods. Next, we review past and recent applications of ML approaches to problems in combustion, spanning fundamental combustion investigations, propulsion and energy-conversion systems, and fire and explosion hazards. Challenges unique to CombML are discussed and further opportunities are identified, focusing on interpretability, uncertainty quantification, robustness, consistency, creation and curation of benchmark data, and the augmentation of ML methods with prior combustion-domain knowledge.

33 ADVANCED PROPULSION SYSTEMS↗

Three-dimensional paganica fault morphology obtained from hypocenter clustering (L'Aquila 2009 seismic sequence, Central Italy)

In seismic modelling, fault planes are normally assumed to be flat due to the lack of data which can constrain fault morphology. However, incorporating 3D fault morphology is important for modelling several phenomena, for example calculating mainshock induced stress changes. Here we utilize a data-analytical method to unveil the 3D rupture morphology of faults using unsupervised clustering techniques applied to earthquake hypocenters in seismic sequences. We apply this method to the 2009 L'Aquila seismic sequence which involved a M W 6.1 mainshock on April 6th. We use a dataset of about 50,000 relocated events, mostly microearthquakes, reaching magnitude of completeness equal to 0.7. Clustering distinguishes the earthquakes as occurring in three main clusters along with other minor fault segments. We then represent the morphology of the main Paganica fault system (responsible for the largest mainshock) using splines. This method shows promise as a step toward robustly and quickly obtaining 3D rupture morphologies where earthquake sequences have been monitored. The 3D model is presented interactively online, and the processing is presented in an interactive Jupyter Notebook (https://bit.ly/2MnCFdj).

58 GEOSCIENCES↗

Discovering phases, phase transitions, and crossovers through unsupervised machine learning: A critical examination

Here, we apply unsupervised machine learning techniques, mainly principal component analysis (PCA), to compare and contrast the phase behavior and phase transitions in several classical spin models - the square and triangular-lattice Ising models, the Blume-Capel model, a highly degenerate biquadratic-exchange spin-one Ising (BSI) model, and the 2D XY model, and examine critically what machine learning is teaching us. We find that quantified principal components from PCA not only allow exploration of different phases and symmetry-breaking, but can distinguish phase transition types and locate critical points. We show that the corresponding weight vectors have a clear physical interpretation, which is particularly interesting in the frustrated models such as the triangular antiferromagnet, where they can point to incipient orders. Unlike the other well-studied models, the properties of the BSI model are less well known. Using both PCA and conventional Monte Carlo analysis, we demonstrate that the BSI model shows an absence of phase transition and macroscopic ground-state degeneracy. The failure to capture the 'charge' correlations (vorticity) in the BSI model (XY model) from raw spin configurations points to some of the limitations of PCA. Finally, we employ a nonlinear unsupervised machine learning procedure, the 'antoencoder method', and demonstrate that it too can be trained to capture phase transitions and critical points.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Online-compatible unsupervised nonresonant anomaly detection

There is a growing need for anomaly detection methods that can broaden the search for new particles in a model-agnostic manner. Most proposals for new methods focus exclusively on signal sensitivity. However, it is not enough to select anomalous events—there must also be a strategy to provide context to the selected events. In this work, we propose the first complete strategy for unsupervised detection of nonresonant anomalies that includes both signal sensitivity and a data-driven method for background estimation. Our technique is built out of two simultaneously trained autoencoders that are forced to be decorrelated from each other. This method can be deployed off-line for nonresonant anomaly detection and is also the first complete on-line-compatible anomaly detection strategy. We show that our method achieves excellent performance on a variety of signals prepared for the ADC2021 data challenge.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Analysis of multispectral data using an unsupervised classification technique: Application to VAS

A statistical classification method based on clustering of multidimensional histograms was applied to several channels of the VAS multispectral imagery. The method automatically discriminates and classifies atmospheric ground features such as cloud types, atmospheric moisture patterns, ocean, or ground. Such a clustering method has the advantage of forming natural data groupings, without a priori classification. Clusters are not limited by straight lines or plane surfaces as is the case in threshold methods. The method was applied to simultaneous full resolution images from channels 8 (11.2 micron), 10 (6.7 micron), and 12 (3.9 micron). Twenty image segments of 64 by 64, 12 image segments of 128 by 128, and 4 image segments of 254 by 254 picture elements were analyzed. In addition, normal VISSR mode images at 1800, 1830, and 2000 GMT were used to identify the classes. The gray levels measured along a scan line and the result of the classification scheme (dashed curves) for the three channels investigated are shown. Each point of the image is affected to a class. Each class is identified by a center of gravity that is represented by a vector in the three dimensional space of gray levels.

Szejwach, G.↗

Notes for the improvement of the spatial and spectral data classification method

This report examines the spatial and spectral clustering technique for the unsupervised automatic classification and mapping of earth resources satellite data, and makes theoretical analysis of the decision rules and tests in order to suggest how the method might best be applied to other flight data such as Skylab and Spacelab.

Dalton, C. C.↗

SDN-NGenIA, a Software Defined Next Generation Integrated Architecture for HEP and Data Intensive Science

This project aims to enable the LHC and other leading programs in high energy physics and other global science domains funded by the DOE to operate with a new level of efficiency and control, through the development of a Next Generation Integrated Architecture (NGenIA) based on intelligent software defined network (SDN)-driven network systems coupled to high throughput applications. While the initial focus is on the challenging LHC use case, the systems and products being developed are general, and apply to many fields of data intensive science ranging from astrophysical sky surveys to bioinformatics and earth observation, as well as other organizations facing the challenges of extracting knowledge from distributed multi-petabyte data stores. A central concept in this development program is a new paradigm ‘consistent network operations’ among widely distributed computing and storage facilities, where stable high throughput flows, at set rates, cross load-balanced network paths, up to flexible high water marks that are adjusted in real time to accommodate other network traffic. The large smooth flows are launched and managed by SDN services that act in concert with the experiments’ site-resident data distribution and management systems, to meet the expanding needs of the science programs. The technical goals include the construction of autonomous, intelligent site-resident services that interact dynamically with network-resident services, and with the science programs’ principal data distribution and management tools, to request or command network resources in support of high throughput petascale to exascale workflow Specific work items include: 1. Developing compact Data Transfer Nodes (DTNs) with auto-tuning functions that support data transfer rates in the 100 Gbps to the 1 Tbps range when used with high throughput data transfer applications 2. Deep site orchestration among virtualized clusters, storage subsystems and subnets to successfully coschedule CPU, storage and network resources 3. Science-program designed site architectures, operational modes, and policy and resource usage priorities, adjudicated across multiple network domains and virtual organizations, using the orchestration functions and methods 4. Seamlessly extending end-to-end operation across both extra-site and intra-site boundaries through the use of Open vSwitch (OVS), FireQoS + next generation Science DMZs 5. Novel methods of system integration that enable granular control of extreme scale long distance transfers through flow matching of scattered source-destination address pairs to multi-domain dynamic circuits 6. Funneling massive sets of streams to DTNs at the site edge hosting petascale buffer pools configured for flows of 100 Gbps and up, exploiting state of the art data transfers 7. Adaptive scheduling based on pervasive end-to-end monitoring, including DTN or compute-node resident agents providing comprehensive end-system profiling 8. Developing unsupervised and supervised machine learning and modeling methods to optimize the workflow involving terabyte to multi-petabyte datasets.

47 OTHER INSTRUMENTATION↗

Machine learning to discover mineral trapping signatures due to CO 2 injection

Mineral trapping is pursued as a geological CO 2 sequestration (GCS) mechanism because it permanently stores CO 2 in solid phases or minerals. However, CO 2 mineral-trapping mechanisms are poorly understood due to (1) lack of sufficient field and laboratory data characterizing these complex processes, and (2) challenges to develop site-specific reactive-transport models coupling fluid flow and geochemical reactions occurring at various temporal (from milliseconds to years) and spatial (from pore (millimeters) to field (kilometers)) scales. Reactive transport with additional complexities such as heterogeneity can make the simulation outputs even more difficult to interpret because of complex nonlinearity and multi-scale interdependencies. Furthermore, the values of model outputs such as concentrations can vary by several orders of magnitude, making it harder to correlate and characterize the impact of the variables via traditional data interpretation techniques such as exploratory data analyses. Recently, machine learning (ML) has shown promise in feature discovery and in highlighting hidden mechanisms that cannot be obtained by existing data-analytics and statistical methods. In this study, we applied an unsupervised ML approach, non-negative matrix factorization with custom -means clustering (NMF) to the data generated by reactive-transport simulations of GCS. The reactive-transport data consisted of 19 attributes, including four physio-chemical variables (pH, porosity, aqueous CO 2 , and sequestered CO 2 ), six chemical species (K + , Na + , HCO, Ca 2+ , Mg 2+ , Fe 2+ ), and four carbonate minerals (calcite, dolomite, siderite, and ankerite), a feldspar mineral (albite), and four clay minerals (illite, clinochlore, kaolinite, and smectite) over a period of 200 years of simulation time. Furthermore, the simulation data used was for Morrow B sandstone at the Farnsworth hydrocarbon unit in Texas. Data are sampled at two locations within the model domain: (1) at the injection well and (2) 200 m west of the injection well. The injection was performed for a period of 10 years. Using NMF, we estimated the temporal interdependencies among the 19 attributes over a span of 200 years. We found that NMF was able to identify four reaction stages and their dominant attributes; these cannot be directly discerned through traditional visualization (e.g., line plots, Pareto analysis, Glyph-based visualization methods) or exploratory data analysis tools of the simulation data. The four stages were: reactions in the injection phase followed by short-, mid-, and long-term reactions. The NMF analysis also revealed that 10 among the 19 attributes are dominant. These dominant attributes for mineral trapping include calcite, dolomite at injection well, siderite at 200 m away from the injection well, clinochlore, kaolinite, Na + , K + , Ca 2+ , Mg 2+ , pH, and aqeuous CO 2 . Finally, at late times (65–200 years), our results showed that calcite plays a major role in mineral trapping with insignificant contribution from siderite, ankerite, and clay minerals. These findings make the proposed unsupervised ML-model attractive for reactive-transport sensing towards real-time GCS monitoring.

54 ENVIRONMENTAL SCIENCES↗

Big Data Analytics for Scanning Transmission Electron Microscopy Ptychography

Electron microscopy is undergoing a transition; from the model of producing only a few micrographs, through the current state where many images and spectra can be digitally recorded, to a new mode where very large volumes of data (movies, ptychographic and multi-dimensional series) can be rapidly obtained. In this paper, we discuss the application of so-called “big-data” methods to high dimensional microscopy data, using unsupervised multivariate statistical techniques, in order to explore salient image features in a specific example of BiFeO 3 domains. Remarkably, k-means clustering reveals domain differentiation despite the fact that the algorithm is purely statistical in nature and does not require any prior information regarding the material, any coexisting phases, or any differentiating structures. While this is a somewhat trivial case, this example signifies the extraction of useful physical and structural information without any prior bias regarding the sample or the instrumental modality. Further interpretation of these types of results may still require human intervention. Finally, however, the open nature of this algorithm and its wide availability, enable broad collaborations and exploratory work necessary to enable efficient data analysis in electron microscopy.

36 MATERIALS SCIENCE↗

Sub-pixel high-resolution imaging of high-energy x-rays inspired by sub-wavelength optical imaging

We have developed and demonstrated an image super-resolution method—XR-UNLOC: X-Ray UNsupervised particle LOCalization—for hard x-rays measured with fast-frame-rate detectors that is an adaptation of the principle of photo-activated localization microscopy (PALM) and stochastic optical reconstruction microscopy (STORM), which enabled biological fluorescence imaging at sub-optical-wavelength scales. We demonstrate the approach on experimental coherent Bragg diffraction data measured with 52 keV x-rays from a nanocrystalline sample. From this sample, we resolve the fine fringe detail of a high-energy x-ray Bragg coherent diffraction pattern to an upsampling factor of 16 of the native pixel pitch of 30 μ m of a charge-integrating fastCCD detector. This was accomplished by analysis of individual photon locations in a series of “nearly-dark” instances of the diffraction pattern that each contain only a handful of photons. Central to our approach was the adaptation of the UNLOC photon fitting routine for PALM/STORM to the hard x-ray regime to handle much smaller point spread functions, which required a different statistical test for photon detection and for sub-pixel localization. A comparison to a photon-localization strategy used in the x-ray community (“droplet analysis”) showed that XR-UNLOC provides significant improvement in super-resolution. We also developed a metric by which to estimate the limit of reliable upsampling with XR-UNLOC under a given set of experimental conditions in terms of the signal-to-noise ratio of a photon detection event and the size of the point spread function for guiding future x-ray experiments in many disciplines where detector pixelation limits must be overcome.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Unsupervised Dictionary Learning for Signal-to-Noise Ratio Enhancement of Array Data

Waveform enhancement methods generally explore lateral coherency in arrivals, often assuming a linear moveout across an array, as exhibited by plane waves. Here, we illustrate how unsupervised dictionary learning combined with orthogonal matching pursuit for feature extraction can be used for signal-to-noise ratio (SNR) enhancement. In this strategy, waveform characteristics are directly learned from provided data samples; the created dictionary is then used for signal extraction. This combination prevents the need to set a predefined dictionary, and it becomes computationally efficient because learning is only done on smaller data portions. Because the dictionary is learned from data, there is no assumption regarding wavefront shape or form. Tests on synthetic and field data demonstrate the better denoising performance in terms of SNR enhancement compared to other methods.

58 GEOSCIENCES↗

ACTINN: automated identification of cell types in single cell RNA sequencing

Abstract Motivation Cell type identification is one of the major goals in single cell RNA sequencing (scRNA-seq). Current methods for assigning cell types typically involve the use of unsupervised clustering, the identification of signature genes in each cluster, followed by a manual lookup of these genes in the literature and databases to assign cell types. However, there are several limitations associated with these approaches, such as unwanted sources of variation that influence clustering and a lack of canonical markers for certain cell types. Here, we present ACTINN (Automated Cell Type Identification using Neural Networks), which employs a neural network with three hidden layers, trains on datasets with predefined cell types and predicts cell types for other datasets based on the trained parameters. Results We trained the neural network on a mouse cell type atlas (Tabula Muris Atlas) and a human immune cell dataset, and used it to predict cell types for mouse leukocytes, human PBMCs and human T cell sub types. The results showed that our neural network is fast and accurate, and should therefore be a useful tool to complement existing scRNA-seq pipelines. Availability and implementation The codes and datasets are available at https://figshare.com/articles/ACTINN/8967116. Tutorial is available at https://github.com/mafeiyang/ACTINN. All codes are implemented in python. Supplementary information Supplementary data are available at Bioinformatics online.

Ma, Feiyang (ORCID:0000000262600787)↗

Physics constrained unsupervised deep learning for rapid, high resolution scanning coherent diffraction reconstruction

By circumventing the resolution limitations of optics, coherent diffractive imaging (CDI) and ptychography are making their way into scientific fields ranging from X-ray imaging to astronomy. Yet, the need for time consuming iterative phase recovery hampers real-time imaging. While supervised deep learning strategies have increased reconstruction speed, they sacrifice image quality. Furthermore, these methods’ demand for extensive labeled training data is experimentally burdensome. Here, we propose an unsupervised physics-informed neural network reconstruction method, PtychoPINN, that retains the factor of 100-to-1000 speedup of deep learning-based reconstruction while improving reconstruction quality by combining the diffraction forward map with real-space constraints from overlapping measurements. In particular, PtychoPINN gains a factor of 4 in linear resolution and an 8 dB improvement in PSNR while also accruing improvements in generalizability and robustness. This blend of performance and computational efficiency offers exciting prospects for high-resolution real-time imaging in high-throughput environments such as X-ray free electron lasers (XFELs) and diffraction-limited light sources.

97 MATHEMATICS AND COMPUTING↗

AUTOMATED UNSUPERVISED CLASSIFICATION OF THE SLOAN DIGITAL SKY SURVEY STELLAR SPECTRA USING k-MEANS CLUSTERING

Large spectroscopic surveys require automated methods of analysis. This paper explores the use of k-means clustering as a tool for automated unsupervised classification of massive stellar spectral catalogs. The classification criteria are defined by the data and the algorithm, with no prior physical framework. We work with a representative set of stellar spectra associated with the Sloan Digital Sky Survey (SDSS) SEGUE and SEGUE-2 programs, which consists of 173,390 spectra from 3800 to 9200 A sampled on 3849 wavelengths. We classify the original spectra as well as the spectra with the continuum removed. The second set only contains spectral lines, and it is less dependent on uncertainties of the flux calibration. The classification of the spectra with continuum renders 16 major classes. Roughly speaking, stars are split according to their colors, with enough finesse to distinguish dwarfs from giants of the same effective temperature, but with difficulties to separate stars with different metallicities. There are classes corresponding to particular MK types, intrinsically blue stars, dust-reddened, stellar systems, and also classes collecting faulty spectra. Overall, there is no one-to-one correspondence between the classes we derive and the MK types. The classification of spectra without continuum renders 13 classes, the color separation is not so sharp, but it distinguishes stars of the same effective temperature and different metallicities. Some classes thus obtained present a fairly small range of physical parameters (200 K in effective temperature, 0.25 dex in surface gravity, and 0.35 dex in metallicity), so that the classification can be used to estimate the main physical parameters of some stars at a minimum computational cost. We also analyze the outliers of the classification. Most of them turn out to be failures of the reduction pipeline, but there are also high redshift QSOs, multiple stellar systems, dust-reddened stars, galaxies, and, finally, odd spectra whose nature we have not deciphered. The template spectra representative of the classes are publicly available in the online journal.

79 ASTRONOMY AND ASTROPHYSICS↗