Datasets:

SouthernCrossAI
/

Arxiv

Dataset card Data Studio Files Files and versions Community

Split (1)

train · 132k rows

Filename stringlengths 22 64	Paragraph stringlengths 8 5.57k
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Proof Sketch. Since DR = DT , by taking expectations we get for the ﬁrst condition: E[ˆeST ] ≤ 0 and hence eMT (I)−δ·eMT ≤ 0. For the second equation we get: E[ˆeSR MT (I) − γˆeSR ] ≤ 0 and hence eMT (I) − eMT (I) − γeMT + γeMT ≤ 0, which implies 0 ≤ 0.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	The ﬁrst term above is analogous to the one in Deﬁni- tion 2.1. The second term captures robustness and asks how representative is the test error of MT (I) w.r.t. its error on other high probability samples when compared with the performance of MT on the same test and robust sets. This can be viewed as an or- thogonal metric to evaluate interpretable procedures in the practical setting. This deﬁnition could also be adapted to a more intuitive but restrictive deﬁnition analogous to Deﬁnition 2.2.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	The second condition vanishes and the ﬁrst condition is just the deﬁnition of δ-interpretability. Our ex- tended deﬁnition is thus consistent with Deﬁnition 2.1 where we have access to the target distribution.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Remark: Model evaluation sometimes requires us to use multiple training and test sets, such as when do- ing cross-validation. In such cases, we have multiple target models M i T trained on independent data sets, and multiple independent test samples Si T (indexed by i = {1, . . . , K}). The empirical error above can be deﬁned as ((cid:80)K T , as well as the training sets, are sampled from the same target distri- bution DT , the reduction to Deﬁnition 2.1 would still apply to this average error, since E[ˆeSi ] ∀ T M h T h, i, j, k.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	We now look at how some of the current methods and how they ﬁt into our framework.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Table 1. Above we see how our framework can be used to characterize interpretability of methods across applications.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	velop a stepwise process for generating selected EDCs based on signiﬁcance testing and broad applicability to various types of expenditures ((Chang & Weiner, 2010), Additional File 1). This stepwise process can be viewed as a (δ, γ)-interpretable procedure that pro- vides information in the form of 19 EDC variables which, when added to the TM, improve the perfor- mance from 103.64% to 95.83% and is thus (0.925, 0) -interpretable. Note the signiﬁcance since, given the large population sizes and high mean annual health- care costs per individual, even small improvements in accuracy can have high monetary impact.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	dataset. Their main objective here is to see if their approach is superior to other approaches in terms of selecting the true important features. There is no robustness experiment here so DR is identity which means same as DT and hence γ is 0. Their accuracy however, in selecting the important features is signif- icantly better than random feature selection which is signiﬁed by δ = 0.1. The other experiments can also be characterized in similar fashion. In cases where only explanations are provided with no explicit metric one can view as the experts conﬁdence in the method as a metric which good explanations will enhance.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Distillation: DNNs are not considered interpretable. However, if you consider a DNN to be a TM then you can view defensive distillation as a (δ, γ)-interpretable procedure. We compute δ and γ from results presented in (Carlini & Wagner, 2017) on the MNIST dataset, where the authors adversarially perturbs the test in- stances by a slight amount. We see here that defensive distillation makes the DNN slightly more robust at the expense of it being a little less accurate.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Prototype Selection: We implemented and ran MMD-critic (Kim et al., 2016) on randomly sampled MNIST training sets of size 1500 where the number of prototypes was set to 200. The test sets were 5420 in size which is the size of the least frequent digit. We had a representative test set and then 10 highly skewed test sets where each contained only a single digit. The representative test set was used to estimate δ and the 10 skewed test sets were used to compute γ. The TM was a nearest prototype classiﬁer (Kim et al., 2016) that was initialized with random 200 prototypes which it used to create the initial classiﬁcations. We see from the table that mmd-critic has a low δ and almost 1 γ value. This implies that it is signiﬁcantly more accurate than random prototype selection while maintaining robustness. In other words, it should be strongly preferred over random selection.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Rule Lists: We built a rule list on a semi-conductor manufacturing dataset (Dhurandhar & Petrik, 2014) In this data, a single datapoint is a of size 8926. wafer, which is a group of chips, and measurements, which correspond to 1000s of input features (temper- atures, pressures, gass ﬂows, etc.), are made on this wafer throughout its production. The goal was to pro- vide the engineer some insight into, if anything, was plaguing his process so that he can improve perfor- mance. We built a rule list (Su et al., 2016) of size at most 4 which we showed to the engineer. The engineer ﬁgured out that there was some issue with some gas ﬂows, which he then ﬁxed. This resulted in 1% more of his wafers ending up within speciﬁcation. Or his yield increased from 80% to 81%. This is signiﬁcant since, even small increase in yield corresponds to billions of dollars in savings.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	It seems that our deﬁnition of δ-interpretability re- quires a predeﬁned goal/task. While (semi-)supervised settings have a well-deﬁned target, we discuss other applicable settings.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	procedure would aim to improve upon. The target distribution in this case would just be the instances in the dataset. If a generative process is assumed for the data, then that would dictate the target distribution.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	In other settings such as reinforcement learning (RL) (Sutton & Barto, 1998), the δ-interpretable procedure would try to increase the expected discounted reward of the agent by directing it into more lucrative portions of the state space.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	or even something else that has superhuman cognitive skills. This generalization allows us to test our deﬁni- tion without the need to pin down human complexity. Moreover, we provide a formal strict deﬁnition for δ- interpretability that accounts for key concepts such as performance and robustness and articulates how ro- bustness is only an issue when talking about incom- plete test sets.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	There maybe situations where a human TM may not want to take any explicit actions based on the insight conveyed by a δ-interpretable procedure. However, an implicit evaluation metric, such as human satisfaction, probably exists, which a good interpretable procedure will help to enhance. For example, in the setting where you want explanations for classiﬁcations of individual instances (Ribeiro et al., 2016), the metric that you are enhancing is the human conﬁdence in the complex model.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	There has been a great deal of interest in interpretable modeling recently and for good reason. Almost any practical application with a human decision maker in- terpretability is imperative for the human to have con- ﬁdence in the model. It has also become increasingly important in deep neural networks given their suscep- tibility to small perturbations that are humanly unrec- ognizable (Carlini & Wagner, 2017; Goodfellow et al., 2016).
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	We deﬁned δ for a single distribution but it could be deﬁned over multiple distributions where δ = max(δ1, ..., δk) for k distributions and analogously γ also could be deﬁned over multiple adversarial distri- butions. We did not include these complexities in the deﬁnitions so as not to lose the main point, but exten- sions such as these are straightforward.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Another extension could be to have a sensitivity pa- rameter α to deﬁne equivalence classes, where if two models are δ1- and δ2-interpretable, then they are in the same equivalence class if \|δ1 − δ2\| ≤ α. This can help group together models that can be considered to be equivalent for the application at hand. The α in essence quantiﬁes operational signiﬁcance. One can have even multiple α as a function of the δ values.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	One can also extend the notion of interpretability where δ or and γ are 1 but you can learn the same model with fewer samples given information from the interpretable procedure.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	There have been multiple frameworks and algorithms proposed to perform interpretable modeling. These range from building rule/decision lists (Wang & Rudin, 2015; Su et al., 2016) to ﬁnding prototypes (Kim et al., 2016) to taking inspiration from psycho- metrics (Id & Dhurandhar, 2017) and learning mod- els that can be consumed by humans. There are also works (Ribeiro et al., 2016) which focus on answering instance speciﬁc user queries by locally approximating a superior performing complex model with a simpler easy to understand one which could be used to gain conﬁdence in the complex model.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Human subjects store approximately 7 pieces of infor- mation (Lisman & Idiart, 1995). As such, we can ex- plore highly interpretable models, which can be readily learned by humans, by considering models for TM that make simple use of no more than 7 pieces of informa- tion. Feldman (Feldman, 2000) ﬁnds that the subjec- tive diﬃculty of a concept is directly proportional to its Boolean complexity, the length of the shortest log- ically equivalent propositional formula. We could ex- plore interpretable models of this type. Another model bounds the rademacher complexity of humans (Zhu et al., 2009) as a function of complexity of the domain and sample size. Although the bounds are loose, they follow the empirical trend seen in their experiments on words and images.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Finally, all humans may not be equal relative to a task. Having expertise in a domain may increase the level of detail consumable by that human. So the above mod- els which try to approximate human capability may be extended to account for the additional complexity consumable by the human depending on their experi- ence.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	We would like to thank Margareta Ackerman, Mur- ray Campbell, Alexandra Olteanu, Marek Petrik, Irina Rish, Kush Varshney and Bowen Zhou for insightful suggestions and comments.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Ribeiro, Marco, Singh, Sameer, and Guestrin, Carlos. ”why should i trust you? explaining the predictions of any classiﬁer. In ACM SIGKDD Intl. Conference on Knowledge Discovery and Data Mining, 2016.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Starﬁeld, B, Weiner, J, Mumford, L, and Steinwachs, D. Ambulatory care groups: a categorization of di- agnoses for research and management. Health Ser- vices Research, 26(5):53–74, 1991.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Carlini, Nicholas and Wagner, David. Towards eval- uating the robustness of neural networks. In IEEE Symposium on Security and Privacy, 2017.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Chang, Hsien-Yen and Weiner, Jonathan P. An in- depth assessment of a diagnosis-based risk adjust- ment model based on national health insurance claimsl the application of the johns hopkins adjusted clinical group case-mix system in taiwan. BMC Medicine, 8(7), 2010.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Su, Guolong, Wei, Dennis, Varshney, Kush R., and Interpretable two-level Malioutov, Dmitry M. Boolean rule learning for classiﬁcation. In Proc. 2016 ICML Workshop Human Interpretability in Machine Learning, pp. 66–70, New York, NY, June 2016.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Dhurandhar, Amit and Petrik, Marek. Eﬃcient and accurate methods for updating generalized linear models with multiple feature additions. Journal of Mach. Learning Research, 2014.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Varshney, Kush R. Engineering safety in machine learning. In Proc. Inf. Theory Appl. Workshop, La Jolla, CA, January–February 2016.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Doshi-Velez, Finale and Kim, Been. Towards a rig- orous science of interpretable machine learning. In https://arxiv.org/abs/1702.08608v2, 2017.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Egele, Manuel, Scholte, Theodoor, Kirda, Engin, and Kruegel, Christopher. A survey on automated dy- namic malware-analysis techniques and tools. ACM Comput. Surv., 44(2):6:1–6:42, 2012.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Zhu, Xiaojin, Gibson, Bryan R, and Rogers, Timo- thy T. Human rademacher complexity. In Ad- vances in Neural Information Processing Systems, pp. 2322–2330, 2009.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Id, Tsuyoshi and Dhurandhar, Amit. Supervised item response models for informative prediction. Knowl. Inf. Syst., 51(1):235–257, April 2017.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Kim, Been, Khanna, Rajiv, and Koyejo, Oluwasanmi. Examples are not enough, learn to criticize! criti- cism for interpretability. In In Advances of Neural Inf. Proc. Systems, 2016.
Processed_A_Formal_Framework_to_Characterize_Interpretabilit.txt	Lisman, John E and Idiart, Marco AP. Storage of 7 plus/minus 2 short-term memories in oscillatory subcycles. Science, 267(5203):1512, 1995.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	Abstract. XMM-Newton observations of the starburst galaxy NGC 1511 reveal the presence of a previously un- known extended hot gaseous phase of its ISM, which partly extends out of the disk plane. The emission distribution is asymmetric, being brightest in the eastern half of the galaxy, where also radio continuum observations suggest the highest level of star formation. Spectral analysis of the integral 0.2–12 keV X-ray emission from NGC 1511 indicates a complex emission composition. A model comprising a power law plus thermal plasma component, both absorbed by foreground gas, cannot explain all details of the observed spectrum, requiring a third spectral component to be added. This component can be a second thermal plasma, but other spectral models can be ﬁtted as well. Its X-ray properties characterize NGC 1511 as a starburst galaxy. The X-ray-to-infrared luminosity ratio is consistent with this result. Together with the X-ray data, XMM-Newton obtained UV images of NGC 1511, trac- ing massive stars heating the ambient gas, which is then seen in Hα emission. UV, Hα and near-infrared imagery suggest that NGC 1511 is disturbed, most likely by its two small companions, NGC 1511a and NGC 1511b.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	The most recent generation of X-ray satellites, XMM- Newton and Chandra, has a much higher sensitivity than previous X-ray missions. This enables us to conduct obser- vations of fainter targets than before and to obtain a much more detailed picture of those known already. Increased sensitivity leads to signiﬁcant progress in investigations of low surface-brightness emitters, such as for example hot gaseous halos around actively star-forming spiral galaxies. As part of the XMM-Newton Guaranteed Time program we have observed the edge-on galaxy NGC 1511. Here we report on the results from these observations.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	The XMM-Newton observations of nearby starburst galaxies form part of a multi-wavelength study of all phases of extraplanar gas in external galaxies, which is conducted in order to assess the importance of halos as repositories of a metal-enriched medium and their signiﬁ- cance in terms of galactic chemical evolution and possible metal enrichment of the intergalactic medium (IGM; e.g.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	Heckman et al. 1990; Dahlem et al. 1995, 1998; Weaver et al. 2000; Heckman 2001). Soft X-ray surveys of nearby spi- ral galaxies, including starburst systems, based on ROSAT data were published e.g. by Read et al. (1997) and Read & Ponman (2001).
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	troscopy was published by Lehnert & Heckman (1995) and, recently, by Kewley et al. (2001), establishing its starburst nature. Two extraplanar gas phases were previ- ously detected in NGC 1511: Hα emission from warm ion- ized gas (Lehnert & Heckman 1995) and nonthermal radio continuum emission from cosmic-rays and magnetic ﬁelds (Dahlem et al. 2001). These detections, together with the high surface brightness of tracers of star formation (SF) in its disk, indicate that it is an actively star-forming galaxy. It is amongst the ﬁrst starburst galaxies observed with XMM-Newton, together with NGC 253 (Pietsch et al. 2001), NGC 3628 and M 82 (unpublished). Chandra obser- vations of NGC 253, M 82 and NGC 3628 were published by Strickland et al. (2000, 2002), Kaaret et al. (2001) and Matsumoto et al. (2001).
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	For an overview of papers related to investigations of galactic gaseous halos, refer to Dettmar (1992) and Dahlem (1997). A description of the XMM-Newton ob- servatory has been given by Jansen et al. (2001).
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	pn and both MOS cameras). No Reﬂection Grating Spectrometer (RGS, den Herder et al. 2001) results are used here, because the target is too faint and too extended for high-resolution X-ray spectroscopy.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	Our observations are the ﬁrst dedicated X-ray observa- tions of NGC 1511. Using the ROSAT All-Sky Survey (RASS) Browser we veriﬁed that it was not detected in the RASS. In the data reduction we followed roughly the procedure adopted by the XMM-Newton Survey Science Centre (SSC, Watson et al. 2001) to split up the total passband into several energy bands. The only diﬀerence is that we chose to split the 0.5–2.0 keV band into two bands, see Table 2. In one case we combined two SSC bands to show an image with energies ranging from 2.0 keV to 7.5 keV.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	Our observations of NGC 1511 were carried out in 2000, July 7–8. The total on-source integration time of the European Photon Imaging Camera (EPIC) pn camera (Str¨uder et al. 2001) is 40.0 ks. The pn camera performed the observation in extended full frame mode and the two MOS cameras (Turner et al. 2001) were put in full frame mode. For all cameras the thin ﬁlter position was chosen, oﬀering the highest possible transmission for very soft X- ray emission. The data were calibrated using the XMM- Newton Science Analysis System (SAS), version 5.3.3, and the Current Calibration File (CCF) version of 2002-10- 10. Good time intervals were deﬁned when the total count rate, including background, was < 1 cts s−1. After removal of data from time intervals with high background count rates a net integration time of 32.1 ks remains (for the pn camera).
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	Corrections were applied to the data for the eﬀects of gain variations of the preampliﬁer channels and charge transfer ineﬃciency (CTI), bad pixels, chip edges, split events etc. and the counts were converted to the pulse- height invariant (PI) energy scale. Also, in order to ﬁll the gaps in the EPIC pn camera images, images in the corresponding energy bands were produced from the data of the two MOS cameras and all images aligned and su- perimposed onto each other. The ﬁnal images presented here are a combination of data from all three EPIC cam- eras. Photons with energies below 0.2 keV were not used in the data analysis because of the calibration uncertain- ties in this very soft energy regime. Derived data products (images and spectra) were produced from the ﬁnal event lists, after application of these corrections.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	In Fig. 1 we display X-ray images of NGC 1511. The photon distributions which were received in the individ- ual energy bands as deﬁned above were smoothed with an adaptive ﬁlter so as to enhance low surface brightness features, while not smearing out the ﬂux of strong point sources. Such images are not suitable for quantitative analyses. The left panel is an overlay of the adaptively- smoothed XMM-Newton 1.3–2.0 keV image on a second generation Digitized Sky Survey (DSS-2) red image. The right panel shows an overlay of the adaptively-smoothed XMM-Newton 2.0–7.5 keV image on the same DSS-2 im- age of NGC 1511. Contours start at the 2.5-σ conﬁdence level above the background and are spaced by factors of √2. The 2.5-σ levels of the images displayed here are: 1.14 10−6 cts s−1 arcmin−2 (0.5–1.3 keV), 0.62 10−6 cts s−1 arcmin−2 (1.3–2.0 keV) and 1.20 10−6 cts s−1 arcmin−2 (2.0–7.5 keV).
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	Fig. 1. XMM-Newton 1.3–2.0 keV image (left), and 2.0–7.5 keV image (right) of NGC 1511 on DSS-2. For details see text. The displayed ﬁeld-of-view is 255′′ 21.6 kpc, for an adopted distance towards NGC 1511 of 17.5 Mpc.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	tre of NGC 1511, which we will call–according to the IAU convention–XMMU J035936.6–673736, does not have an optical counterpart in either our XMM-Newton Optical Monitor (OM, Mason et al. 2001) images or the DSS. If associated with NGC 1511, it might be an ultra-luminous (LX = 1.18 1040 erg s−1) X-ray source (ULX, see e.g. Strickland et al. 2001, Foschini et al. 2002 and references therein). If not related to NGC 1511, it could be a back- ground source, e.g. an AGN (see Sect. 3.2.4).
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	determine the local background.1 Extracted spectra were grouped based both on a minimum number of counts (25) for each channel and on the maximum number of spec- tral channels sampling the pn energy resolution (3). An analysis of the background-subtracted integral spectrum of NGC 1511 over the energy range 0.2–12.0 keV was per- formed using Xspec v. 11.1.0.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	clearly requires a multi-component ﬁt. However, the one presented here is not necessarily the only possible good ﬁt to the spectrum (Table 3). At this time, we can only state that a single thermal component plus power law spectral model cannot explain the entire soft X-ray spectrum of NGC 1511, demonstrating the complexity of the emission spectrum. The nature of the additional third component is as yet uncertain.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	pared to other galaxies, is thus relatively low (Read & Ponman 2001; their Fig. 5, central panel). The tempera- tures of the thermal plasma components are in line with earlier measurements of other starburst galaxies, based on ROSAT and ASCA data (e.g. Dahlem et al. 1998; Weaver et al. 2000 and references therein). Components at about 0.2–0.3 keV and 0.6 keV were found by us in several star- burst galaxies, with the 0.2–0.3 keV component dominant in the halos, while the hotter component is normally ob- served from gas in the disks. Similar results were also ob- tained based on previously published data from Chandra and XMM-Newton (see references in Sect. 1).
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	Fig. 2. Total XMM-Newton EPIC pn spectrum of NGC 1511. The spectral model ﬁtted to the data includes two thermal plasma components plus a power law, all with absorption by foreground gas (model “MMP” in Table 3). In the lower panel the residuals after subtraction of the spectral model are presented.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	and an absorbing column density of NH = 2.5+1.0 −0.3 1021 cm−2. The 0.2–12 keV X-ray ﬂux of this source is fX = 3.22 10−13 erg s−1 cm−2. The ﬁtted moderately low ab- sorbing column density makes it unlikely that this source is a background AGN, which one would expect to have much higher intrinsic absorption.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	The OM onboard XMM-Newton, with its three UV ﬁlters in addition to a range of optical ﬁlters, opens a second window on the targets observed that is not accessible from the ground, in the UV. As an example, we present in the left panel of Fig. 3 a 230 nm image of NGC 1511 obtained with the UVM2 ﬁlter.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	This image exhibits the UV emission from massive stars, which are the primary sources of energy heating the warm ionized gas in NGC 1511. This is the reason for the global correspondence between the emission max- ima in the UV and Hα images. However, it is clear from a comparison of the two frames (Fig. 3 [left panel] and Fig. 4 [right panel]) that there is more diﬀuse Hα outside the H II-regions. The presence of such massive stars is yet another indicator of NGC 1511’s activity, as already sug- gested by its LX/LFIR ratio and also by the presence of diﬀuse halo emission in various wavebands.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	In the overlay with the EPIC 0.5–1.3 keV soft X-ray image in the right panel of Fig. 3 one can see that there is basically no spatial correspondence between UV and X-ray emitters in NGC 1511.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	The UV point source south-east of the centre of NGC 1511 is probably an unrelated foreground star. It also shows up on an R-band image by Lehnert & Heckman (1995). Its position does not coincide with that of SN1935C or Nova Hydri 1935 (van den Bergh & Hazen 1988).
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	The Hα frame shows quite impressively how strong the starburst activity in NGC 1511 is and also how disturbed the galaxy is. This disturbance is most likely caused by its companions, NGC 1511a and NGC 1511b, which are observed at projected distances from NGC 1511 of 56.7 kpc and 38.7 kpc, respectively, with very small relative recession velocities of only 38(25) km s−1 with respect to NGC 1511. Using a high-contrast display of the cor- responding red DSS-2 plate one can see a tidal tail con- nected to NGC 1511 (not visible in Fig. 1, because with such a grey-scale no contours could be seen), from the south-eastern end of the galaxy pointing roughly north- wards, suggesting an ongoing tidal interaction.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	An OM image obtained with the UVW2 ﬁlter (cen- tral wavelength about 210 nm), exhibits the same gen- eral emission distribution as Fig. 3, however with a lower signal-to-noise ratio. Therefore, it is not displayed here.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	The observations described here are part of a large project aiming at understanding the nature of gaseous halos around actively star-formating galaxies.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	In the case of NGC 1511 it is interesting to compare the XMM-Newton images with others obtained in diﬀer- ent wavebands. Especially a comparison of the OM image with Hα and 2 Micron All-Sky Survey (2MASS; Skrutskie 2001) near-infrared frames is useful. Both, the Hα frame by Lehnert & Heckman (1995) and the 2MASS J-H-Ks- band composite image are displayed in Fig. 4.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	The starburst nature of NGC 1511 is also corroborated by an updated radio spectral index calculation that we performed, using our earlier measurements (Dahlem et al. 2001), plus a 4.85 GHz measurement by Gregory et al. (1994) and an 843 MHz ﬂux density measurement from the Sydney University Molonglo Sky Survey (SUMSS; Bock et al. 1999). S4.85GHz = 76 mJy and S0.843GHz = 237 0.08 mJy lead to a new, improved value of ¯α = ν ¯α), a value typical of actively star-forming galaxies (S (Condon 1992). This new value is, within the ample error bar of our earlier calculations, consistent with these. The Molonglo Telescope 843 MHz map (not displayed) also corroborates the emission distribution of our 1.43 GHz and 2.45 GHz maps (Dahlem et al. 2001) with a low major-to- minor axial ratio, indicating the presence of a synchrotron radio halo.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	The X-ray disk emission follows roughly that of Hα and ra- dio continuum (Lehnert & Heckman 1995; Dahlem et al.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	of the dependence of its gaseous halos’ properties on the level and distribution of star formation in the underlying disk, for which one would prefer to study relatively undis- turbed systems. Processes other than energy input from stellar sources (stellar winds and supernovae; see Leitherer & Heckman 1995) might inﬂuence the gas motions verti- cal to the disk, such as e.g. a gravitational potential mod- iﬁed by a tidal interaction with a companion galaxy. Still, NGC 1511 is one of a small number of starburst galax- ies in which various components of the ISM are observed, within the same volume, above their disk planes. A signif- icant contribution to the heating of its ISM from a hidden AGN can be excluded based on our X-ray data due to the absence of a hard central source, from optical spectra because of the absence of typical AGN emission lines and from our radio data, which also do not show any signatures of AGN activity at the centre.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	Bock D. C.-J., Large M. I. & Sadler E. M. 1999, AJ, 117, 1578 Condon J. J. 1992, ARA&A, 30, 575 Dahlem M. 1997, PASP, 109, 1298 Dahlem M., Lazendic J., Haynes R. F., Ehle M. & Lisenfeld U.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	NGC 1511, amongst other objects, is also being observed by us in H I line emission with the Australia Telescope Compact Array to study the properties of its neutral atomic gas. This will lead to new information about the intrinsic absorbing gas distribution and its kinematics. It will, for example, also help to elucidate the issue whether the observed disturbances of NGC 1511’s disk can be ex- plained by interactions with its companions, NGC 1511a and NGC 1511b.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	XMM-Newton observations of several more nearby starburst galaxies, including NGC 1808, NGC 3628 and NGC 4666, have been conducted by us. These, together with data from the XMM-Newton archive, will be used to study systematically and in a self-consistent manner the soft X-ray emission properties of galaxies with kpc- to galaxy-scale outﬂows from their disks.
Processed_The_quest_for_hot_gas_in_the_halo_of_NGC_1511.txt	Acknowledgements: We thank Dr. M. Lehnert for making available to us his Hα and R-band images of NGC 1511 in digital form. This work is based on observations obtained with XMM-Newton, an ESA science mission with instru- ments and contributions directly funded by ESA Member States and the USA (NASA). This research has made use of the NASA/IPAC Infrared Science Archive, which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. The Digitized Sky Surveys were produced at the Space Telescope Science Institute under U.S. Government grant NAG W-2166. The images of these surveys are based on photographic data obtained using the Oschin Schmidt Telescope on Palomar Mountain and the UK Schmidt Telescope. The plates were processed into the present compressed digital form with the permission of these institutions.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	Abstract— Understanding crowd mobility behaviors would be a key enabler for crowd management in smart cities, beneﬁting various sectors such as public safety, tourism and transportation. This article discusses the existing challenges and the recent advances to overcome them and allow sharing information across stakeholders of crowd management through Internet of Things (IoT) technologies. The article proposes the usage of the new federated interoperable semantic IoT platform (FIESTA-IoT), which is considered as “a system of systems”. The platform can support various IoT applications for crowd management in smart cities. In particular, the article discusses two integrated IoT systems for crowd mobility: 1) Crowd Mobility Analytics System, 2) Crowd Counting and Location System (from the SmartSantander testbed). Pilot studies are conducted in Gold Coast, Australia and Santander, Spain to fulﬁll various requirements such as providing online and ofﬂine crowd mobility analyses with various sensors in different regions. The analyses provided by these systems are shared across applications in order to provide insights and support crowd management in smart city environments.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	Sustainable development of cities is a major global chal- lenge as more than half of the world population is living in urban areas. The smart city concept allows optimizing services for urban areas because or as a result of the advancement of the new technologies ranging from very small devices to big data centers. These technologies can be considered in the context of IoT, where many objects, devices, machines, and data centers are connected. The usage of IoT technologies for crowd management in urban environments is promising for the future of smart cities.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	IoT technologies can enable many improvements for crowd management, which spans sectors such as transportation ser- vices (e.g., operating public transport or directing pedestrian trafﬁc), public safety (e.g., detection of ﬁghting incidents), and tourism (e.g., event management for enhanced visitor experience). For instance, movement behaviors of crowds may indicate situations such as trafﬁc congestion, emergency incidents, and panic situations during certain events such as large gatherings in city squares.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	Fig. 1. Federated and interoperable IoT platform supporting crowd manage- ment stakeholders of smart cities.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	learned insights is the key technical enabler to understand the crowd mobility behaviors in a smart city. Finally, privacy preservation is a problem of utmost importance for smart cities. While various data from vast deployment of sensors travel through the IoT systems, preserving privacy at a level closer to the data contributors (providers) is an important challenge.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	This article describes the recent advances in IoT for un- derstanding crowd mobility in smart cities. The federated and interoperable semantic IoT (FIESTA-IoT) platform for smart cities is introduced for the speciﬁc perspective of crowd management applications. Fig. 1 illustrates the outlook of the smart city applications leveraging the smart city platform for sharing information across various stakeholders. While the platform is currently in use for several smart city testbeds, the article focuses on two IoT systems for crowd mobility, namely Crowd Mobility Analytics System (CMAS) and Crowd Counting and Location System (CCLS) and discusses the aspects related to the aforementioned limitations.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	Two pilot studies are conducted in Gold Coast, Australia and Santander, Spain, where various sensors are deployed in urban areas. The ﬁrst pilot study uses CMAS in Gold Coast for a medium-scale smart city deployment. The requirements of the pilot include analyzing heavy or light pedestrian trafﬁc at streets with or without vehicles. The second pilot study uses CCLS in an indoor market in Santander. The requirements in- clude detecting people (crowd size) and locating their positions at public buildings of a city and other critical infrastructures. In both pilots, data anonymisation limits tracking devices for long time periods. On the other hand, online and ofﬂine analytics information needs to be shared across various stakeholders such as city councils and visualized in several interfaces using IoT technologies and infrastructure to provide insights for crowd management in smart cities.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	Smart city data is often gathered by solutions where dedi- cated networks of sensors or data sources produce observations to be consumed by speciﬁc applications. The systems usually differ from each other, serving for distinct purposes, and they are mostly not interoperable [1], [2]. In this regard, creating crowd management services that harness the abundant data from a smart city (e.g., environmental data, road trafﬁc infor- mation) would require either ad-hoc integration or creation of new systems. This situation raises a new requirement of an integrated “system of systems” or “container of systems”.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	FIESTA-IoT systems that can be leveraged. Currently more than 5000 sensors (from 11 integrated testbeds [4]) report environmental data (e.g., temperature, humidity, illuminance, noise level), road trafﬁc information (e.g., vehicle speed, trafﬁc intensity), car and bike parking spots, estimated arrival times of buses, and smart building information (e.g., human occupancy, power consumption).
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	At the Federated Cloud Infrastructure (middle layer), the data from the bottom layer is modelled using the FIESTA- IoT Semantic Model and stored in the Linked-Data Storage. In particular, the semantic model for crowd mobility data is described in Section II-B. The data in the Cloud infrastructure is accessible through the Federated Context Management which exposes NGSI and SPARQL interfaces. Our open source IoT Broker (Aeron Broker) component provides scalable fed- eration for the context management, whereas IoT Discovery (NEConfMan) enables easy registration and discovery of re- sources with features such as geo-discovery.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	results (generated by CMAS, CCLS or other IoT services) from the Cloud and provide visualizations.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	In order to provide seamless interoperability and informa- tion transparency from IoT systems to the crowd management applications, the crowd mobility outcomes are semantically annotated following the FIESTA-IoT ontology [5] as shown in Fig. 3 (with a stress on the speciﬁc taxonomy of M3-lite for crowd mobility).
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	Rich and complex knowledge is represented with an ontol- ogy as things are connected to each other through relation- ships. Things are not identiﬁed as individuals, but as classes of individuals. Moreover, a class might have sub-classes. For example, peopleCounterX is an instance of PeopleCountSensor class which is a subclass of Counter (see Fig. 3). The classes can be deﬁned and described in taxonomies and an ontology may use classes from different ontologies or taxonomies. Relationships between classes are known as properties and it is possible to deﬁne properties’ cardinality. Each class and property used in an ontology is uniquely identiﬁed by a namespace preﬁx and the class or property speciﬁc name. For example, m3-lite:PeopleCountSensor is a class deﬁned in the M3-lite ontology. For the sake of readability, in this paragraph we are omitting the namespace preﬁx while they are shown with preﬁx in Fig. 3.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	The core concept is the SensingDevice, representing a sensor that produces Observation, which is a measurement (or com- putation) of a phenomenon related to an object happened at a speciﬁc Instant. For example, a crowd mobility detector can be seen as a Device composed of multiple SensingDevices. In this sense, such a detector can have one PeopleFlowCountSensor and one StayingPeopleCountSensor, which are subclasses of PeopleCountSensor. The Observation(s) is expressed with a QuantityKind having a Unit. Following our example, the QuantityKind associated to the data generated by the People- FlowCountSensor is CountPeopleMoving (subclass of Quanti- tyKind) with Item as its Unit and with the Direction property expressed either in geodetic DirectionAzimuth or as a generic DirectionHeading. The directions start from the Point that is the location of the physical Platform. Platform is meant as the supporting dock to which the Device is attached. The Staying- PeopleCountSensor generates CountPeopleStaying values ex- pressed in Item. The system also consists of PeopleStayDura- tionSensor that generates PeopleStayDurationAverage values measured in SecondTime. Each SensingDevice might have a Coverage, speciﬁed either as Polygon/Rectangle/Circle or as a simple Point. This indicates the geographic extent of the Observation.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	Based on the captured Wi-Fi probes, the system can count the mobile devices in these sensing areas. The cameras are co- located with speciﬁc Wi-Fi sniffers deployed at the dedicated calibration choke points. A built-in people counting software runs in the cameras. Both Wi-Fi device detection and people counting results are reported to to the Cloud, where data analytics modules reside, through the IoT gateways. Three analytics modules are developed: crowd estimation, people ﬂows, and stay duration. The crowd estimation module outputs number of people by correlating the stereoscopic camera counts and the number of Wi-Fi enabled devices at the calibration points. Based on the correlation between the two data modalities, the calibration of the data analytical results are applied in other sensing areas without cameras. The module monitoring people ﬂows infers crowd movement in these areas. Finally, the stay duration module estimates the waiting times and the number of waiting people. All analytics results are exported to the Federated Cloud Infrastructure so the crowd analytics results are discoverable and available for applications in the smart city platform.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	The purpose of the data collection from terminal equip- ment is restricted to mere statistical counting.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	The tracking is limited in time and space to the extent strictly necessary for this purpose.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	experimentation, the only option is to anonymise data re- garding to MAC addresses. Thus, experimentation security measures must be undertaken to address both, data integrity and anonymisation. Therefore, any type of experimentation or service provision must take into account this concern, which is usually underestimated by system developers.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	in [10]) or longer travels within the city, but it is the price that must be paid to meet the privacy requirements.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	The deployments target two regions. These sensors deployed in these areas are considered as Cluster 1 for (expected) heavy pedestrian trafﬁc and Cluster 2 for light trafﬁc places. Each cluster has a stereoscopic camera for the calibration. The collected data is sent to the Cloud where two virtual machines are created for the clusters. Clustering the areas allows applying CMAS to city-scale by sharing the raw data load.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	Fig. 5. Weekly measurements from Gold Coast. Cluster 1: Pedestrian ways with heavy pedestrian trafﬁc, Cluster 2: Roads including vehicles and light pedestrian trafﬁc.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	Figure 5-a shows the average number of Wi-Fi devices detected for one-week period. There exists an increased activ- ity in Cluster 1 region especially during Friday (23/03/2018) and the following weekend. This can be due to crowdedness in the shopping street and the beach area contained in this region. Moreover, there is a peak in Saturday that can be due to an event or gathering. Figure 5-b shows the change of the coefﬁcients (ratios). The ratios are computed at the calibration choke points (providing near-ground truth for the measurements). The hourly ratio is computed such that num- ber of people count-in and count-out events are divided to the number of Wi-Fi probes. First, for Cluster 2 with light trafﬁc, correlation coefﬁcient is mostly (almost all days) higher compared to Cluster 1. Second, correlation coefﬁcient values lie mostly in the range of (0.2, 2), whereas the peak value is about 2.8. This indicates that the results based on Wi-Fi-only measurements are likely to have less accuracy most of the times of the days and the correlation changes throughout the days. Lastly, there exists certain regularity in the correlation from one day to another, which can be learned through a time period and then applied to other time periods where camera is temporarily inactive or removed. On the other hand, as seen in the peak hours of Cluster 2, the ratios do not lie within a narrow range. One reason can be events affecting the volume of pedestrians. Lastly, Fig. 5-b shows relatively higher variance of the coefﬁcient for areas with light pedestrian trafﬁc. Calibration could be necessary for shorter time intervals.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	Overall, it is observed that effective use of Wi-Fi sensing and combining them with sensing by stereoscopic cameras produce accurate sensing in large scale for both the heavy and light pedestrian trafﬁc areas. Moreover, the variance between heavy and light trafﬁc shows the usefulness of the clustering approach which treats these regions separately.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	There are recent studies that focus on understanding of human mobility through IoT devices such as wireless sensors. Jara et al. [12] observed the relation between trafﬁc behavior and temperature conditions as a smart city application through deployment of IoT devices in Santander. Tong et al. [13] propose usage of Wi-Fi sensors to understand passenger ﬂows. Evaluation through simulation results shows high accuracy. Zhao et al. [14] survey the recent advances in understanding human mobility in urban environments. The study lists some of the existing urban human mobility datasets collected such as GPS, GSM, Wi-Fi, and Bluetooth traces. Similarly, Zhou et al. [15] discuss the topic of human mobility in urban environments and present a taxonomy of crowdsensed input data types and application outcomes such as crowd density and ﬂows within building, and people transportation mode identiﬁcation (cycling, running, bus riding). Lastly, Montjoye et al. [10] focus on the privacy aspect by analyzing long period Wi-Fi traces and show that 95% of the individuals can be uniquely identiﬁed using spatiotemporal datasets.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	The current work focuses on ﬁnding insights behind crowd mobility such as detecting crowdedness. However, under- standing more complex crowd mobility behaviour in a large- scale city area such as movements of groups (e.g., family) could be helpful for crowd management and enhancing smart mobility in the cities. The collected mobility information can serve as input of human mobility simulations to further study how city dynamics are affected by crowd mobility patterns. With the combination of real mobility dataset in a simulated environment, learning new mobility insights opens up new opportunities for new crowd management strategies (e.g., con- gestion avoidance, evacuation planning, demand management) that can further improve the public service and safety in smart cities.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	In our future developments, the semantic interoperability through ontologies can be leveraged more extensively for cross-infrastructure communication and knowledge sharing. The new advancements of the NGSI protocol by the ETSI Industry Speciﬁcation Group (ISG) on Context Information Management (CIM) are centered around the concepts of linked data. This opens a new horizon where knowledge graphs are shared among various infrastructures and, while their adminis- trators own the produced data, it is still accessible seamlessly and transparently by all actors in the multi-infrastructure federation.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	This article discusses the new advancements towards under- standing crowd mobility in smart cities using IoT. While there exist certain limitations, the CMAS and CCLS systems using the smart city platform offer improvements for more efﬁcient crowd management. The pilot studies in Gold Coast and Santander show the capability to fulﬁll various requirements and share information across stakeholders by leveraging the IoT technologies and infrastructure.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	Acknowledgment: The pilot study in Gold Coast is conducted in collaboration with NEC Australia. This work has been partially funded by the Spanish Government (MINECO) under Grant Agree- ment No. TEC2015-71329-C2-1-R ADVICE (Dynamic Pro- visioning of Connectivity in High Density 5G Wireless Sce- narios) project and by the EU Horizon 2020 Programme un- der Grant Agreements No. 731993 AUTOPILOT (Automated Driving Progressed by Internet Of Things), 643943 FIESTA- IoT (Federated Interoperable Semantic IoT Testbeds and Ap- plications), and 643275 FESTIVAL (Federated Interoperable Smart ICT Services Development and Testing Platforms) projects and the joint project by NEC Laboratories Europe and Technische Universit¨at Dortmund. The content of this paper does not reﬂect the ofﬁcial opinion of the Spanish Government or European Union. Responsibility for the information and views expressed therein lies entirely with the authors.
Processed_Toward_Understanding_Crowd_Mobility_in_Smart_Citie.txt	G ¨URKAN SOLMAZ ([email protected]) is a senior researcher at NEC Laboratories Europe, Germany. His re- search interests are in mobile computing, human mobility analytics, and Internet of Things. FANG-JING WU ([email protected]) is a junior professor at TU Dortmund. Her research interests are pri- marily in pervasive computing, wireless communications and networks, and Internet of Things. FLAVIO CIRILLO (ﬂ[email protected]) is a research scientist at NEC Laboratories Europe. IoT platforms are his main research focus, with a particular interest to scalability, interoperability, and federation aspects. ERN ¨O KOVACS ([email protected]) is the senior man- ager for the IoT Research Group, NEC Laboratories Europe. His context brokering, Cloud-Edge Computing, real-time sit- uation awareness, knowledge extraction, and smart cities. JUAN RAM ´ON SANTANA ([email protected]) is a senior research fellow at the University of Cantabria. Among his research interests are wireless sensor networks, M2M communications, and mobile phone application research. LUIS S ´ACHEZ ([email protected]) is a professor at the Department of Communications Engineering, University of Cantabria. He is active on IoT-enabled smart cities, meshed networking on heterogeneous wireless scenarios, and opti- mization of network performance through cognitive network- ing techniques. PABLO SOTRES ([email protected]) is a senior re- search fellow at the University of Cantabria, Spain. He has been involved in several international projects framed under the smart city paradigm, such as SmartSantander; and related to inter-testbed federation, such as Fed4FIRE, Fed4FIRE+ and Wise-IoT. LUIS MU ˜NOZ ([email protected]) is the head of the Network Planning and Mobile Communications Laboratory at the University of Cantabria, Spain. His research focuses on advanced data transmission techniques, heterogeneous wire- less multihop networks, and applied mathematical methods for telecommunications.
Processed_The_first_high-resolution_observations_of_37.7-,_3.txt	ABSTRACT We have used the Australia Telescope Compact Array (ATCA) to undertake the ﬁrst high angular resolution observations of 37.7-GHz (7−2 → 8−1E) methanol masers towards a sample of eleven high-mass star formation regions which host strong 6.7- GHz methanol masers. The 37.7-GHz methanol sites are coincident to within the astrometric uncertainty (0.4 arcseconds) with the 6.7-GHz methanol masers associated with the same star formation region. However, spatial and spectral comparison of the 6.7- and 37.7-GHz maser emission within individual sources shows that the 37.7-GHz masers are less often, or to a lesser degree co-spatial than are the 12.2-GHz and 6.7- GHz masers. We also made sensitive, high angular resolution observations of the 38.3- and 62 → 53A+, respectively) and 38.5-GHz class II methanol transitions (62 → 53A and the 36.2-GHz (4−1 → 30E) class I methanol transition towards the same sample of eleven sources. The 37.7-, 38.3- and 38.5-GHz methanol masers are unresolved in the current observations, which implies a lower limit on the brightness temperature of the strongest masers of more than 106 K. We detected the 38.3-GHz methanol transition towards 7 sources, 5 of which are new detections and detected the 38.5-GHz transition towards 6 sources, 4 of which are new detections. We detected 36.2-GHz class I methanol masers towards all eleven sources, 6 of these are new detections for this transition, of which 4 sources do not have previously reported class I methanol masers from any transition.

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.