Classification of intermediate redshift ($z$ = 0.3--0.8) emission line galaxies as star-forming galaxies, composite galaxies, active galactic nuclei (AGN), or low-ionization nuclear emission regions (LINERs) using optical spectra alone was impossible because the lines used for standard optical diagnostic diagrams: [NII], H$\alpha$, and [SII] are redshifted out of the observed wavelength range. In this work, we address this problem using four supervised machine learning classification algorithms: $k$-nearest neighbors (KNN), support vector classifier (SVC), random forest (RF), and a multi-layer perceptron (MLP) neural network. For input features, we use properties that can be measured from optical galaxy spectra out to $z < 0.8$---[OIII]/H$\beta$, [OII]/H$\beta$, [OIII] line width, and stellar velocity dispersion---and four colors ($u-g$, $g-r$, $r-i$, and $i-z$) corrected to $z=0.1$. The labels for the low redshift emission line galaxy training set are determined using standard optical diagnostic diagrams. RF has the best area under curve (AUC) score for classifying all four galaxy types, meaning highest distinguishing power. Both the AUC scores and accuracies of the other algorithms are ordered as MLP$>$SVC$>$KNN. The classification accuracies with all eight features (and the four spectroscopically-determined features only) are 93.4% (92.3%) for star-forming galaxies, 69.4% (63.7%) for composite galaxies, 71.8% (67.3%) for AGNs, and 65.7% (60.8%) for LINERs. The stacked spectrum of galaxies of the same type as determined by optical diagnostic diagrams at low redshift and RF at intermediate redshift are broadly consistent. Our publicly available code (https://github.com/zkdtc/MLC_ELGs) and trained models will be instrumental for classifying emission line galaxies in upcoming wide-field spectroscopic surveys.

Twenty years have passed since first light for the Sloan Digital Sky Survey (SDSS). Here, we release data taken by the fourth phase of SDSS (SDSS-IV) across its first three years of operation (July 2014-July 2017). This is the third data release for SDSS-IV, and the fifteenth from SDSS (Data Release Fifteen; DR15). New data come from MaNGA - we release 4824 datacubes, as well as the first stellar spectra in the MaNGA Stellar Library (MaStar), the first set of survey-supported analysis products (e.g. stellar and gas kinematics, emission line, and other maps) from the MaNGA Data Analysis Pipeline (DAP), and a new data visualisation and access tool we call "Marvin". The next data release, DR16, will include new data from both APOGEE-2 and eBOSS; those surveys release no new data here, but we document updates and corrections to their data processing pipelines. The release is cumulative; it also includes the most recent reductions and calibrations of all data taken by SDSS since first light. In this paper we describe the location and format of the data and tools and cite technical references describing how it was obtained and processed. The SDSS website (www.sdss.org) has also been updated, providing links to data downloads, tutorials and examples of data use. While SDSS-IV will continue to collect astronomical data until 2020, and will be followed by SDSS-V (2020-2025), we end this paper by describing plans to ensure the sustainability of the SDSS data archive for many years beyond the collection of data.

To shed light on the time evolution of local star formation episodes in M33, we study the association between 566 Giant Molecular Clouds (GMCs), identified through the CO (J=2-1) IRAM-all-disk survey, and 630 Young Stellar Cluster Candidates (YSCCs), selected via Spitzer-24~$\mu$m emission. The spatial correlation between YSCCs and GMCs is extremely strong, with a typical separation of 17~pc, less than half the CO(2--1) beamsize, illustrating the remarkable physical link between the two populations. GMCs and YSCCs follow the HI filaments, except in the outermost regions where the survey finds fewer GMCs than YSCCs, likely due to undetected, low CO-luminosity clouds. The GMCs have masses between 2$\times 10^4$ and 2$\times 10^6$ M$_\odot$ and are classified according to different cloud evolutionary stages: inactive clouds are 32$\%$ of the total, classified clouds with embedded and exposed star formation are 16$\%$ and 52$\%$ of the total respectively. Across the regular southern spiral arm, inactive clouds are preferentially located in the inner part of the arm, possibly suggesting a triggering of star formation as the cloud crosses the arm. Some YSCCs are embedded star-forming sites while the majority have GALEX-UV and H$\alpha$ counterparts with estimated cluster masses and ages. The distribution of the non-embedded YSCC ages peaks around 5~Myrs with only a few being as old as 8--10~Myrs. These age estimates together with the number of GMCs in the various evolutionary stages lead us to conclude that 14~Myrs is a typical lifetime of a GMC in M33, prior to cloud dispersal. The inactive and embedded phases are short, lasting about 4 and 2~Myrs respectively. This underlines that embedded YSCCs rapidly break out from the clouds and become partially visible in H$\alpha$ or UV long before cloud dispersal.

We present Herschel/HIFI observations (WISH KP) of 14 water lines in a small sample of galactic massive protostellar objects: NGC6334I(N), DR21(OH), IRAS16272-4837, and IRAS05358+3543. We analyze the gas dynamics from the line profiles. Through modeling of the observations using RATRAN, we estimate outflow, infall, turbulent velocities, molecular abundances, and investigate any correlation with the evolutionary status of each source. The molecular line profiles exhibit a broad component coming from the shocks along the cavity walls associated with the protostars, and an infalling (or expansion for IRAS05358+3543) and passively heated envelope component, with highly supersonic turbulence likely increasing with the distance from the center. Accretion rates between 6.3 10^{-5} and 5.6 10^{-4} \msun yr^{-1} are derived from the infall observed in three of our sources. The outer water abundance is estimated to be at the typical value of a few 10^{-8} while the inner abundance varies from 1.7 10^{-6} to 1.4 10^{-4} with respect to H2 depending on the source. We confirm that regions of massive star formation are highly turbulent and that the turbulence likely increases in the envelope with the distance to the star. The inner abundances are lower than the expected 10^{-4} perhaps because our observed lines do not probe deep enough into the inner envelope, or because photodissociation through protostellar UV photons is more efficient than expected. We show that the higher the infall/expansion velocity in the protostellar envelope, the higher is the inner abundance, maybe indicating that larger infall/expansion velocities generate shocks that will sputter water from the ice mantles of dust grains in the inner region. High-velocity water must be formed in the gas-phase from shocked material.