-
Greenery Segmentation In Urban Images By Deep Learning
Authors:
Artur A. M. Oliveira,
Nina S. T. Hirata,
Roberto Hirata Jr
Abstract:
Vegetation is a relevant feature in the urban scenery and its awareness can be measured in an image by the Green View Index (GVI). Previous approaches to estimate the GVI were based upon heuristics image processing approaches and recently by deep learning networks (DLN). By leveraging some recent DLN architectures tuned to the image segmentation problem and exploiting a weighting strategy in the l…
▽ More
Vegetation is a relevant feature in the urban scenery and its awareness can be measured in an image by the Green View Index (GVI). Previous approaches to estimate the GVI were based upon heuristics image processing approaches and recently by deep learning networks (DLN). By leveraging some recent DLN architectures tuned to the image segmentation problem and exploiting a weighting strategy in the loss function (LF) we improved previously reported results in similar datasets.
△ Less
Submitted 12 December, 2019;
originally announced December 2019.
-
Kernels on fuzzy sets: an overview
Authors:
Jorge Guevara,
Roberto Hirata Jr,
Stéphane Canu
Abstract:
This paper introduces the concept of kernels on fuzzy sets as a similarity measure for $[0,1]$-valued functions, a.k.a. \emph{membership functions of fuzzy sets}.
We defined the following classes of kernels: the cross product, the intersection, the non-singleton and the distance-based kernels on fuzzy sets.
Applicability of those kernels are on machine learning and data science tasks where unc…
▽ More
This paper introduces the concept of kernels on fuzzy sets as a similarity measure for $[0,1]$-valued functions, a.k.a. \emph{membership functions of fuzzy sets}.
We defined the following classes of kernels: the cross product, the intersection, the non-singleton and the distance-based kernels on fuzzy sets.
Applicability of those kernels are on machine learning and data science tasks where uncertainty in data has an ontic or epistemistic interpretation.
△ Less
Submitted 30 July, 2019;
originally announced July 2019.
-
A logical-based corpus for cross-lingual evaluation
Authors:
Felipe Salvatore,
Marcelo Finger,
Roberto Hirata Jr
Abstract:
At present, different deep learning models are presenting high accuracy on popular inference datasets such as SNLI, MNLI, and SciTail. However, there are different indicators that those datasets can be exploited by using some simple linguistic patterns. This fact poses difficulties to our understanding of the actual capacity of machine learning models to solve the complex task of textual inference…
▽ More
At present, different deep learning models are presenting high accuracy on popular inference datasets such as SNLI, MNLI, and SciTail. However, there are different indicators that those datasets can be exploited by using some simple linguistic patterns. This fact poses difficulties to our understanding of the actual capacity of machine learning models to solve the complex task of textual inference. We propose a new set of syntactic tasks focused on contradiction detection that require specific capacities over linguistic logical forms such as: Boolean coordination, quantifiers, definite description, and counting operators. We evaluate two kinds of deep learning models that implicitly exploit language structure: recurrent models and the Transformer network BERT. We show that although BERT is clearly more efficient to generalize over most logical forms, there is space for improvement when dealing with counting operators. Since the syntactic tasks can be implemented in different languages, we show a successful case of cross-lingual transfer learning between English and Portuguese.
△ Less
Submitted 23 October, 2019; v1 submitted 10 May, 2019;
originally announced May 2019.
-
Concave losses for robust dictionary learning
Authors:
Rafael Will M de Araujo,
Roberto Hirata,
Alain Rakotomamonjy
Abstract:
Traditional dictionary learning methods are based on quadratic convex loss function and thus are sensitive to outliers. In this paper, we propose a generic framework for robust dictionary learning based on concave losses. We provide results on composition of concave functions, notably regarding super-gradient computations, that are key for develo** generic dictionary learning algorithms applicab…
▽ More
Traditional dictionary learning methods are based on quadratic convex loss function and thus are sensitive to outliers. In this paper, we propose a generic framework for robust dictionary learning based on concave losses. We provide results on composition of concave functions, notably regarding super-gradient computations, that are key for develo** generic dictionary learning algorithms applicable to smooth and non-smooth losses. In order to improve identification of outliers, we introduce an initialization heuristic based on undercomplete dictionary learning. Experimental results using synthetic and real data demonstrate that our method is able to better detect outliers, is capable of generating better dictionaries, outperforming state-of-the-art methods such as K-SVD and LC-KSVD.
△ Less
Submitted 2 November, 2017;
originally announced November 2017.
-
maigesPack: A Computational Environment for Microarray Data Analysis
Authors:
Gustavo H. Esteves,
Roberto Hirata Jr
Abstract:
Microarray technology is still an important way to assess gene expression in molecular biology, mainly because it measures expression profiles for thousands of genes simultaneously, what makes this technology a good option for some studies focused on systems biology. One of its main problem is complexity of experimental procedure, presenting several sources of variability, hindering statistical mo…
▽ More
Microarray technology is still an important way to assess gene expression in molecular biology, mainly because it measures expression profiles for thousands of genes simultaneously, what makes this technology a good option for some studies focused on systems biology. One of its main problem is complexity of experimental procedure, presenting several sources of variability, hindering statistical modeling. So far, there is no standard protocol for generation and evaluation of microarray data. To mitigate the analysis process this paper presents an R package, named maigesPack, that helps with data organization. Besides that, it makes data analysis process more robust, reliable and reproducible. Also, maigesPack aggregates several data analysis procedures reported in literature, for instance: cluster analysis, differential expression, supervised classifiers, relevance networks and functional classification of gene groups or gene networks.
△ Less
Submitted 11 November, 2015; v1 submitted 19 September, 2014;
originally announced September 2014.
-
Lithium, Carbon, and Oxygen Abundances of Hyades F-G Type Stars
Authors:
Yoichi Takeda,
Satoshi Honda,
Takashi Ohnishi,
Michiko Ohkubo,
Ryuko Hirata,
Kozo Sadakane
Abstract:
In an attempt to carry out a systematic study on the behavior of the photospheric abundances of Li, C, and O (along with Fe) for Hyades main-sequence stars in the T_eff range of ~5000-7000K, we conducted an extensive spectrum-synthesis analysis applied to four spectral regions (comprising lines of Fe-group elements, Li I 6708 line, C I 7111-7119 lines, and O I 6156-8 lines) based on the high-dispe…
▽ More
In an attempt to carry out a systematic study on the behavior of the photospheric abundances of Li, C, and O (along with Fe) for Hyades main-sequence stars in the T_eff range of ~5000-7000K, we conducted an extensive spectrum-synthesis analysis applied to four spectral regions (comprising lines of Fe-group elements, Li I 6708 line, C I 7111-7119 lines, and O I 6156-8 lines) based on the high-dispersion spectra of 68 selected F-G type stars belonging to this cluster. The abundances of C and O turned out to be fairly uniform in a marginally supersolar level such like the case of Fe: <[C/H]> = +0.15 (sigma = 0.08), <[O/H]> = +0.22 (sigma = 0.14), and <[Fe/H]> = +0.11(sigma = 0.08), suggesting that the primordial abundances are almost retained for these elements. Strictly, however, they show a slightly increasing trend with a decrease in T_eff (typically on the order of ~10^(-4) dex/K; while this might be due to an improper choice of atmospheric parameters, we found it hard to give a quantitatively reasonable explanation. Regarding Li, we confirmed the well-known T_eff-dependent trend in the Li abundance reported so far (a conspicuous Li-trough at 6300K <T_eff< 6700K and a progressive decrease toward a lower T_eff at T_eff < 6000K), which means that the surface Li of Hyades stars is essentially controlled only by T_eff and other parameters such as the rotational velocity are almost irrelevant.
△ Less
Submitted 27 December, 2012;
originally announced December 2012.
-
Multisite spectroscopic seismic study of the beta Cep star V2052 Oph: inhibition of mixing by its magnetic field
Authors:
M. Briquet,
C. Neiner,
C. Aerts,
T. Morel,
S. Mathis,
D. R. Reese,
H. Lehmann,
R. Costero,
J. Echevarria,
G. Handler,
E. Kambe,
R. Hirata,
S. Masuda,
D. Wright,
S. Yang,
O. Pintado,
D. Mkrtichian,
B. -C. Lee,
I. Han,
A. Bruch,
P. De Cat,
K. Uytterhoeven,
K. Lefever,
J. Vanautgaerden,
B. de Batz
, et al. (7 additional authors not shown)
Abstract:
We used extensive ground-based multisite and archival spectroscopy to derive observational constraints for a seismic modelling of the magnetic beta Cep star V2052 Ophiuchi. The line-profile variability is dominated by a radial mode (f_1=7.14846 d^{-1}) and by rotational modulation (P_rot=3.638833 d). Two non-radial low-amplitude modes (f_2=7.75603 d^{-1} and f_3=6.82308 d^{-1}) are also detected.…
▽ More
We used extensive ground-based multisite and archival spectroscopy to derive observational constraints for a seismic modelling of the magnetic beta Cep star V2052 Ophiuchi. The line-profile variability is dominated by a radial mode (f_1=7.14846 d^{-1}) and by rotational modulation (P_rot=3.638833 d). Two non-radial low-amplitude modes (f_2=7.75603 d^{-1} and f_3=6.82308 d^{-1}) are also detected. The four periodicities that we found are the same as the ones discovered from a companion multisite photometric campaign (Handler et al. 2012) and known in the literature. Using the photometric constraints on the degrees l of the pulsation modes, we show that both f_2 and f_3 are prograde modes with (l,m)=(4,2) or (4,3). These results allowed us to deduce ranges for the mass (M \in [8.2,9.6] M_o) and central hydrogen abundance (X_c \in [0.25,0.32]) of V2052 Oph, to identify the radial orders n_1=1, n_2=-3 and n_3=-2, and to derive an equatorial rotation velocity v_eq \in [71,75] km s^{-1}. The model parameters are in full agreement with the effective temperature and surface gravity deduced from spectroscopy. Only models with no or mild core overshooting (alpha_ov \in [0,0.15] local pressure scale heights) can account for the observed properties. Such a low overshooting is opposite to our previous modelling results for the non-magnetic beta Cep star theta Oph having very similar parameters, except for a slower surface rotation rate. We discuss whether this result can be explained by the presence of a magnetic field in V2052 Oph that inhibits mixing in its interior.
△ Less
Submitted 21 August, 2012;
originally announced August 2012.
-
Non-LTE Calculation for the Be Star Decretion Disk
Authors:
Hidetoshi Iwamatsu,
Ryuko Hirata
Abstract:
The non-LTE state of the hydrogen gas in isothermal transonic decretion disks around a B1V star has been calculated by an iterative method in order to explore the basic physical process in the disk. This dynamical model is characterized by a density law in the equatorial plane of $ρ(R) \propto R^{-3.5}$. The continuous radiation is calculated with the $Λ$ iteration in the integral form, while we…
▽ More
The non-LTE state of the hydrogen gas in isothermal transonic decretion disks around a B1V star has been calculated by an iterative method in order to explore the basic physical process in the disk. This dynamical model is characterized by a density law in the equatorial plane of $ρ(R) \propto R^{-3.5}$. The continuous radiation is calculated with the $Λ$ iteration in the integral form, while we adopt a single-flight escape probability for lines. We describe the non-LTE state, the radiation flow and conversion in the disk. We conclude that the stellar Balmer continuum plays a key role in the non-LTE state of the disk. The examination of the local energy gain and loss suggests that the disk temperature has double minima along the equatorial plane in the optically thick case: the intermediate region caused by deficient ultraviolet radiation and the outer Lyman $α$ cooling region. We have also calculated some observable quantities, such as the spectral energy distribution, the $UBV$ colors, the infrared excess and the Balmer line profiles. Our calculations with the mass loss rate less than $10^{-10} M_{\odot} {\rm yr}^{-1}$ reproduce the observed continuum quantities. However, we could not get large H$α$ emission strength observed in Be stars. We suggest that the density gradient of the Be star disk is slower than that of the isothermal decretion disk.
△ Less
Submitted 21 May, 2008;
originally announced May 2008.
-
HV Virginis and WZ Sge-Type Dwarf Novae
Authors:
Taichi Kato,
Yoshiyuki Sekine,
Ryuko Hirata
Abstract:
A dwarf nova HV Vir was observed photometrically for eight nights during the outburst in 1992 April - May. The star showed two distinct types of periodic variation: (1) 82.20-min (0.05708 d) double-humped variation with decaying amplitudes during the early stage of the outburst, and (2) 83.80-min (0.05820 d) superhumps in later stages. We attributed the former to "early superhumps", which are on…
▽ More
A dwarf nova HV Vir was observed photometrically for eight nights during the outburst in 1992 April - May. The star showed two distinct types of periodic variation: (1) 82.20-min (0.05708 d) double-humped variation with decaying amplitudes during the early stage of the outburst, and (2) 83.80-min (0.05820 d) superhumps in later stages. We attributed the former to "early superhumps", which are only seen in the earliest stage of WZ Sge-type outbursts. The superhump period and evolution of the superhumps, together with general characteristics of the light curve, make HV Vir a typical WZ Sge-type dwarf nova. HV Vir also showed a large increase of the superhump period during the superoutburst. Upon the recognition of the WZ Sge-type nature of an object previously considered as a nova, we present a comprehensive list of candidates for WZ Sge-type dwarf novae, and related systems.
△ Less
Submitted 9 October, 2001;
originally announced October 2001.
