Search

A key yet unresolved question in modern-day astronomy is how galaxies formed and evolved under the paradigm of the $\Lambda$CDM model. A critical limiting factor lies in the lack of robust tools to describe the merger history through a statistical model. In this work, we employ a generative graph network, E(n) Equivariant Graph Normalizing Flows Model. We demonstrate that, by treating the progenitors as a graph, our model robustly recovers their distributions, including their masses, merging redshifts and pairwise distances at redshift z=2 conditioned on their z=0 properties. The generative nature of the model enables other downstream tasks, including likelihood-free inference, detecting anomalies and identifying subtle correlations of progenitor features., Comment: 6 pages, 3 figures, accepted to the ICML 2022 Machine Learning for Astrophysics workshop

Chemical kinetics plays an important role in governing the thermal evolution in reactive flows problems. The possible interactions between chemical species increase drastically with the number of species considered in the system. Various ways have been proposed before to simplify chemical networks with an aim to reduce the computational complexity of the chemical network. These techniques oftentimes require domain-knowledge experts to handcraftedly identify important reaction pathways and possible simplifications. Here, we propose a combination of autoencoder and neural ordinary differential equation to model the temporal evolution of chemical kinetics in a reduced subspace. We demonstrated that our model has achieved a close-to 10-fold speed-up compared to commonly used astro-chemistry solver for a 9-species primordial network, while maintaining 1 percent accuracy across a wide-range of density and temperature., Comment: 10 pages, 8 figures, accepted to the ICML 2022 Machine Learning for Astrophysics workshop

Although automated pathology classification using deep learning (DL) has proved to be predictively efficient, DL methods are found to be data and compute cost intensive. In this work, we aim to reduce DL training costs by pre-training a ResNet feature extractor using SimCLR contrastive loss for latent encoding of OCT images. We propose a novel active learning framework that identifies a minimal sub-sampled dataset containing the most uncertain OCT image samples using label propagation on the SimCLR latent encodings. The pre-trained ResNet model is then fine-tuned with the labelled minimal sub-sampled data and the underlying pathological sites are visually explained. Our framework identifies upto 2% of OCT images to be most uncertain that need prioritized specialist attention and that can fine-tune a ResNet model to achieve upto 97% classification accuracy. The proposed method can be extended to other medical images to minimize prediction costs.

Web browsers have evolved into indispensable software platforms for modern applications. Although JavaScript has become a de-facto lingua franca for developing web applications, the JavaScript software development tools and methodologies lag behind those of languages such as Java and C#. In particular, to enhance JavaScript programs with non-functional concerns (e.g., persistence, security, transactions, and logging) JavaScript developers need to modify the source code by hand. Developers then often end up having to maintain the resulting modified version separately for reasons that include poor design, the complexities of program evolution, and fundamental language limitations. The necessity to modify code by hand could be avoided if JavaScript had metadata facilities for developers to specify non-functional concerns declaratively. To address this problem, we present a novel declarative approach, T ransparent A utomated E nhancement for J ava S cript (TAE-JS) that enhances JavaScript programs with the ability to use declarative metadata. The metadata is expressed by means of Java annotations. We have implemented our technique in an open-source plug-in for the Eclipse IDE. This plug-in allows a developer to enhance the original version of a JavaScript program by specifying concerns declaratively using Java annotations. It then analyzes an original version of the program to automatically generate enhanced program versions by adding the declared concerns. We evaluated TAE-JS with four practical examples that enhance JavaScript programs by declaratively adding concerns such as persistence, security, transactions, and logging. In a user study with ten student developers and three professional engineers at PayPal, the participants viewed favorably TAE-JS's flexible declarative enhancement capabilities, seeing them as a valuable mechanism for implementing non-functional concerns in JavaScript programs. Our evaluation results are promising in demonstrating the potential benefits of our approach to complement existing JavaScript development tools.

As the world wide web grows rapidly, a text corpus is becoming increased online at an incredible rate. Managing a corpus of documents is critical for many areas of science, industry, and culture. For example, bioengineering researchers, who study a new generation of advanced materials, frequently need to identify and understand a comprehensive body of literature describing an association between material features of interest. However, there is no inspection technique to help such researchers who need to make critical decisions based on their understanding of a corpus of documents. In this paper, we present a text visualization approach, TOPEXPLORER which extracts and visualizes topic models regarding bioengineering document collections. TOPEXPLORER displays text data in a logical layout to inspect and understand the relations among documents. It applies three probabilistic topic modeling algorithms as complementary study methods to systematically discover hidden thematic structures in a collection of documents. In the evaluation, we assessed TOPEXPLORER by building topic models on 600 documents in the bioengineering domain. The interactive visualization of TOPEXPLORER allows users to explore the hidden structure that a topic model discovers. TOPEXPLORER helps users understand and explore the output of models by effectively organizing, summarizing, visualizing, and interacting with a corpus.

Developers often make recurring changes, similar but different changes across multiple locations. They inspect such code changes per source file (i.e., a diff patch) during code reviews; however, diff patches represent low-level code modification without summarizing recurring changes, leading to tedious and error-prone code inspection. To address this problem, we propose a novel code review approach, Recurring Code Changes Inspection with Deep Learning (RIDL) that leverages change patterns of an edit script by learning code clones, identical or nearly similar code fragments. To train a classifier, RIDL learns 13,940 clones with four different clone types (e.g., Type-1, Type-2, Type-3, and Type-4 clones) from a clone database mined from 25,000 subject programs. Our approach then leverages the classifier to (1) interactively summarize recurring changes and (2) detect change mistakes, potential anomalies in a given codebase. In the evaluation, after 2 hours of training, RIDL analyzes code changes in four open source projects. It summarizes recurring changes with 95.1% accuracy and detects change anomalies with 93.1% accuracy. Our results show that RIDL should help developers effectively inspect recurring changes during code reviews.

Similar changes are common during the development of a project. Many of the changes are similar but appears different based on the local context. During code review the code changes are inspected per each source file. The process of identifying similar code change is time-taking and error-prone. To overcome this problem, we propose Similar Changes Inspection with Deep Learning (SIL) which (1) creates a generalized edit script based on the data and control dependence to (2) identify and summarize similar code changes by (3) modelling a deep learning classifier. In order to train a classifier, we have identified clones of four types from a clone database mined from 25,000 programs. SIL summarizes the changes and identifies the change anomalies. To obtain feedback on the SIL approach, we have conducted an user study with seven Computer Science students. The study revealed that SIL helped these developers to conduct peer code reviews more effectively. SIL is available as an Eclipse plug-in and its demonstration video is available at https://sites.google.com/unomaha.edu/codereview-deeplearning.

The late stages of stellar evolution from asymptotic giant branch stars to planetary nebulae are now known to be an active phase of molecular synthesis. Over 80 gas-phase molecules have been detected through rotational transitions in the mm/submm region. Infrared spectroscopy has also detected inorganic minerals, fullerenes, and organic solids. The synthesis of these molecules and solids take place over very low density ($, 22 pages, 10 figures

An unidentified infrared emission (UIE) feature at 6.0 $��$m is detected in a number of astronomical sources showing the UIE bands. In contrast to the previous suggestion that this band is due to C=O vibrational modes, we suggest that the 6.0 $��$m feature arises from olefinic double-bond functional groups. These groups are likely to be attached to aromatic rings which are responsible for the major UIE bands. The possibility that the formation of these functional groups is related to the hydrogenation process is discussed., 43 pages, 15 figures, accepted for publication in the Astrophysical Journal

The linewidth of ions has been observed to be systematically narrower than that of the coexisting neutrals in molecular clouds (Houde et al. 2000) and been interpreted as the signature of the decoupling of the neutral turbulence from magnetic fields in partially ionized medium (Li & Houde 2008; Paper I, hereafter). As a sequel of Paper I, here we present further observational evidence that lend support to these earlier proposals with the velocity coordinate spectrum analysis (Lazarian & Pogosyan 2006). We recover the turbulent energy spectra of HCN and HCO+(4-3) in a starless molecular filament in NGC 6334 where magnetic fields play a dynamically important role (Li et al. 2015). Our analysis showed that the neutral spectrum is consistent with Kolmogorov-type (k^-5/3, where k is the wave number), while that of the ions is the same on the large scale but steeper (k^-2) for scales smaller than 0.404pc. We carefully ruled out the possibilities that the spectrum difference can stem from the differences of ion and neutral optical depth and hyper-fine structures.

The question whether magnetic fields play an important role in the processes of molecular cloud and star formation has been debated for decades. Recent observations have revealed a simple picture that may help illuminate these questions: magnetic fields have a tendency to preserve their orientation at all scales that have been probed - from 100-pc scale inter-cloud media down to sub-pc scale cloud cores. This ordered morphology has implications for the way in which self-gravity and turbulence interact with magnetic fields: both gravitational contraction and turbulent velocities should be anisotropic, due to the influence of dynamically important magnetic fields. Such anisotropy is now observed. Here we review these recent observations and discuss how they can improve our understanding of cloud/star formation., Comment: Accepted for publication as a chapter in Protostars and Planets VI, University of Arizona Press (2014), eds. H. Beuther, R. Klessen, C. Dullemond, Th. Henning

The question whether magnetic fields play an important role in the processes of molecular cloud and star formation has been debated for decades. Recent observations have revealed a simple picture that may help illuminate these questions: magnetic fields have a tendency to preserve their orientation at all scales that have been probed - from 100-pc scale inter-cloud media down to sub-pc scale cloud cores. This ordered morphology has implications for the way in which self-gravity and turbulence interact with magnetic fields: both gravitational contraction and turbulent velocities should be anisotropic, due to the influence of dynamically important magnetic fields. Such anisotropy is now observed. Here we review these recent observations and discuss how they can improve our understanding of cloud/star formation.

Recent infrared spectroscopic observations have shown that proto-planetary nebulae (PPNs) are sites of active synthesis of organic compounds in the late stages of stellar evolution. This paper presents a study of Spitzer/IRS spectra for a sample of carbon-rich PPNs, all except one of which show the unidentified 21 micron emission feature. The strengths of the aromatic infrared band (AIB), 21 micron, and 30 micron features are obtained by decomposition of the spectra. The observed variations in the strengths and peak wavelengths of the features support the model that the newly synthesized organic compounds gradually change from aliphatic to aromatic characteristics as stars evolve from PPNs to planetary nebulae., Comment: 32 pages, 11 figures, accepted for publication in ApJ

We present results of our search for collimated ionized winds in bipolar nebulae using the Very Large Array (VLA) and the Australia Telescope Compact Array (ATCA). Our search is motivated by the discovery of an ionized jet in the bipolar nebula M 2-9 (Lim & Kwok 2003) that may be responsible for sculpting the nebula's mirror-symmetric structure. To determine if such jets are a common feature of bipolar nebulae, we searched for optically-thick radio cores - a characteristic signature of ionized jets - in 11 northern nebulae with the VLA at 1.3 cm and 0.7 cm, and in 5 southern nebulae with the ATCA at 6 cm and 3.6 cm. Two northern objects, 19W32 and M 1-91, and two southern objects, He2-84 and and Mz 3, exhibit a compact radio core with a rising spectrum consistent with an ionized jet. Th 2-B exhibits a steeply falling spectrum characteristic of nonthermal radio emission. Here we present a preliminary analysis of these five radio cores and discuss the implications of our results., 4 pages, 1 figure; to appear in Asymmetric Planetary Nebulae III, eds. M. Meixner, J. Kastner, N.Soker and B. Balick (ASP Conf. Ser.)

22 pags., 8 figs., 1 tab., Massive stars disrupt their natal molecular cloud material through radiative and mechanical feedback processes. These processes have profound effects on the evolution of interstellar matter in our Galaxy and throughout the universe, from the era of vigorous star formation at redshifts of 1-3 to the present day. The dominant feedback processes can be probed by observations of the Photo-Dissociation Regions (PDRs) where the far-ultraviolet photons of massive stars create warm regions of gas and dust in the neutral atomic and molecular gas. PDR emission provides a unique tool to study in detail the physical and chemical processes that are relevant for most of the mass in inter-and circumstellar media including diffuse clouds, proto-planetary disks, and molecular cloud surfaces, globules, planetary nebulae, and star-forming regions. PDR emission dominates the infrared (IR) spectra of star-forming galaxies. Most of the Galactic and extragalactic observations obtained with the James Webb Space Telescope (JWST) will therefore arise in PDR emission. In this paper we present an Early Release Science program using the MIRI, NIRSpec, and NIRCam instruments dedicated to the observations of an emblematic and nearby PDR: the Orion Bar. These early JWST observations will provide template data sets designed to identify key PDR characteristics in JWST observations. These data will serve to benchmark PDR models and extend them into the JWST era. We also present the Science-Enabling products that we will provide to the community. These template data sets and Science-Enabling products will guide the preparation of future proposals on star-forming regions in our Galaxy and beyond and will facilitate data analysis and interpretation of forthcoming JWST observations., Support for JWST-ERS program ID 1288 was provided through grants from the STScI under NASA contract NAS5-03127 to STScI (K.G., D.V.D.P., M.R.), Univ. of Maryland (M.W., M.P.), Univ. of Michigan (E.B., F.A.), and Univ. of Toledo (T.S.-Y.L.). O.B. and E.H. are supported by the Programme National “Physique et Chimie du Milieu Interstellaire” (PCMI) of CNRS/INSU with INC/INP co-funded by CEA and CNES, and through APR grants 6315 and 6410 provided by CNES. E. P. and J.C. acknowledge support from the National Science and Engineering Council of Canada (NSERC) Discovery Grant program (RGPIN-2020-06434 and RGPIN-2021-04197 respectively). E.P. acknowledges support from a Western Strategic Support Accelerator Grant (ROLA ID 0000050636). J.R.G. and S.C. thank the Spanish MCINN for funding support under grant PID2019-106110GB-I00. Work by M.R. and Y.O. is carried out within the Collaborative Research Centre 956, subproject C1, funded by the Deutsche Forschungsgemeinschaft (DFG)—project ID 184018867. T.O. acknowledges support from JSPS Bilateral Program, grant No. 120219939. M.P. and M.W. acknowledge support from NASA Astrophysics Data Analysis Program award #80NSSC19K0573. C.B. is grateful for an appointment at NASA Ames Research Center through the San José State University Research Foundation (NNX17AJ88A) and acknowledges support from the Internal Scientist Funding Model (ISFM) Directed Work Package at NASA Ames titled: “Laboratory Astrophysics—The NASA Ames PAH IR Spectroscopic Database.”