source
stringlengths 1
2.05k
⌀ | target
stringlengths 1
11.7k
|
|---|---|
LW deconvolution with subsequent local PSF fitting allows us to recover the positions of point sources with a standard deviation <0.04. pixels. except for the faintest sources. | LW deconvolution with subsequent local PSF fitting allows us to recover the positions of point sources with a standard deviation $<0.04$ pixels, except for the faintest sources. |
Local PSF fitting without deconvolution and fitting with a single PSF lead to results that are of similar quality. | Local PSF fitting without deconvolution and fitting with a single PSF lead to results that are of similar quality. |
Lucy-Richardson deconvolution clearly deteriorates the astrometry. the standard deviations of the stellar positions being 2-3 times higher than in the other cases. | Lucy-Richardson deconvolution clearly deteriorates the astrometry, the standard deviations of the stellar positions being 2-3 times higher than in the other cases. |
We note that the mean of the differences between mpu= and recovered positions ts different from zero i1 all cases (by a few 1/100 of a pixel). | We note that the mean of the differences between input and recovered positions is different from zero in all cases (by a few $1/100$ of a pixel). |
I propose that the most mmportant factor for this behavior ts that the position of a star has been defined in the simulated images to coincide with the centroid position of the PSF (this ts the usual standard for PSF fitting algorithms). | I propose that the most important factor for this behavior is that the position of a star has been defined in the simulated images to coincide with the centroid position of the PSF (this is the usual standard for PSF fitting algorithms). |
In the artificial image. the stars are fixed at their locations and convolved with the complete PSF. | In the artificial image, the stars are fixed at their locations and convolved with the complete PSF. |
StarFinder (like DAOPHOT) fits the positions and fluxes of the sources only within a fitting radius (or a fitting box in the case of StarFinder) and uses the complete PSF only for point source subtraction. | StarFinder (like DAOPHOT) fits the positions and fluxes of the sources only within a fitting radius (or a fitting box in the case of StarFinder) and uses the complete PSF only for point source subtraction. |
The centroid of the PSF within the fitting radius (box) will differ from that of the entire PSF if the PSF outside the fitting radius is not point-symmetrical. | The centroid of the PSF within the fitting radius (box) will differ from that of the entire PSF if the PSF outside the fitting radius is not point-symmetrical. |
That is. the difference in mean positions between input and measurement ts due to a difference between the PSFs used for input (entire model PSF) and output measurement (a partial PSP or a deconvolved PSF). | That is, the difference in mean positions between input and measurement is due to a difference between the PSFs used for input (entire model PSF) and output measurement (a partial PSF or a deconvolved PSF). |
In the case of a single PSF. it appears that there are actually distributions of the differences between input and output. one centered on a slightly positive position. another one centered on slightly negative values. | In the case of a single PSF, it appears that there are actually distributions of the differences between input and output, one centered on a slightly positive position, another one centered on slightly negative values. |
This i$ not necessarily surprising because the PSF becomes elongated with distance from the guide star. which may change the centroid position. depending on which side of the guide star a star is located. | This is not necessarily surprising because the PSF becomes elongated with distance from the guide star, which may change the centroid position, depending on which side of the guide star a star is located. |
However. I did not fully explore this possibility because it would go far beyond the scope of this work if [ were to explore the phenomenon of deviation between input and measured positions 1n detail here. | However, I did not fully explore this possibility because it would go far beyond the scope of this work if I were to explore the phenomenon of deviation between input and measured positions in detail here. |
However. I believe it is important to mention these points here because they may become of great significance in work that requires extremely precise astrometry in AO images. | However, I believe it is important to mention these points here because they may become of great significance in work that requires extremely precise astrometry in AO images. |
It may not be sufficient to determine stellar positions just from the PSF cores. | It may not be sufficient to determine stellar positions just from the PSF cores. |
At the moment. also I cannot exclude the possibility that positions may become biased by deconvolution. depending on the location in the field. | At the moment, also I cannot exclude the possibility that positions may become biased by deconvolution, depending on the location in the field. |
The differences between input and recovered positions for the different methods are shown in refFig:photo.. | The differences between input and recovered positions for the different methods are shown in \\ref{Fig:photo}. |
Both PSF fitting with a single PSF and PSF fitting after LR deconvolution produce significant deviations with the latter method providing the poorer results. | Both PSF fitting with a single PSF and PSF fitting after LR deconvolution produce significant deviations with the latter method providing the poorer results. |
An explanation of the bad performance of the LR algorithm can probably be found in its non-linearity. | An explanation of the bad performance of the LR algorithm can probably be found in its non-linearity. |
The LR deconvolution tends to be influenced by local noise peaks and incorporates the smooth background into the point sources. | The LR deconvolution tends to be influenced by local noise peaks and incorporates the smooth background into the point sources. |
This leads to an increasing overestimation of the flux of faint sources and to characteristic empty patches with a size similar to the PSF around bright sources. | This leads to an increasing overestimation of the flux of faint sources and to characteristic empty patches with a size similar to the PSF around bright sources. |
Both local PSF fitting and local PSF fitting after linear deconvolution provide acceptable results. | Both local PSF fitting and local PSF fitting after linear deconvolution provide acceptable results. |
Local PSF fitting after Wiener deconvolution leads to the smallest standard and mean deviations. | Local PSF fitting after Wiener deconvolution leads to the smallest standard and mean deviations. |
We note that the distribution for the single PSF (upper right panel in refFig:photo)) appears bivariate. | We note that the distribution for the single PSF (upper right panel in \\ref{Fig:photo}) ) appears bivariate. |
This may be related to the PSF properties discussed in the previous paragraph. | This may be related to the PSF properties discussed in the previous paragraph. |
Positive deviations. Le.. a source that is brighter in the measurement than in the input. are largely excluded in the case of a single PSF used for measurement. | Positive deviations, i.e., a source that is brighter in the measurement than in the input, are largely excluded in the case of a single PSF used for measurement. |
Elongation of the sources with distance from the guide star will lead to a loss of flux. | Elongation of the sources with distance from the guide star will lead to a loss of flux. |
Smooth maps of the differences between input and measured photometry across the FOV are shown in refFig:dmag.. | Smooth maps of the differences between input and measured photometry across the FOV are shown in \\ref{Fig:dmag}. |
Using a single PSF leads to a systematic error that increases with distance from the guide star up to - mmag. | Using a single PSF leads to a systematic error that increases with distance from the guide star up to $\sim0.25$ mag. |
LR deconvolution combined with local PSF fitting does not lead to acceptable results. as we have already seen in | LR deconvolution combined with local PSF fitting does not lead to acceptable results, as we have already seen in |
Their approach was further demonstrated to improve the overall accuracy of the mixture modeling solution. | Their approach was further demonstrated to improve the overall accuracy of the mixture modeling solution. |
In this work. we consider the problem of galaxy classilication. based on sky survey data. wilh several unknown classes present in (he data. | In this work, we consider the problem of galaxy classification, based on sky survey data, with several unknown classes present in the data. |
For this domain. we evaluate both and a new approach which we propose here. one (hat is applicable to class discovery lor neural network-basecl (NN) classifiers. | For this domain, we evaluate both \citet{Browning} and a new approach which we propose here, one that is applicable to class discovery for neural network-based (NN) classifiers. |
In section 2. we describe the data sets ancl data preparation. | In section 2, we describe the data sets and data preparation. |
In section 3. we review Miller&Browning(2003a).. MillerBrowning(2003b) and also introduce a class discovery. approach for NN classifiers. | In section 3, we review \citet{pami}, \citet{Browning} and also introduce a class discovery approach for NN classifiers. |
In section 3. we also describe several performance criteria. each capturing different aspects of the class discovery problem. | In section 3, we also describe several performance criteria, each capturing different aspects of the class discovery problem. |
In section 4. we present our experimental results. | In section 4, we present our experimental results. |
Finally. the paper concludes with a summary ancl some ciscussion. | Finally, the paper concludes with a summary and some discussion. |
In our experiments we used (wo data sets. each with over 5000 data points. | In our experiments we used two data sets, each with over 5000 data points. |
The first was data from Storrie-Lombardietal.(1992) (hencelorth denoted as ESOLV after the ESO-LV catalog of Lauberts&Valentijià (1989))) which has been used previously in several studies of automated classification methods (Storrie-Lombarclietal.1992:Owens&Aha 2001). | The first was data from \citet{storrie92} (henceforth denoted as ESOLV after the ESO-LV catalog of \citet{lauberts89}) ) which has been used previously in several studies of automated classification methods \citep{storrie92, owens96, bazell01}. |
. The second data set consisted of SDSS early release data (2002).. composed of over 50000 objects of various (vpes. | The second data set consisted of SDSS early release data \citet{stoughton}, composed of over 50000 objects of various types. |
Slorrie-Lombarclietal.(1992). performed one of the earliest attempts al morphological Classification of galaxies using neural networks. | \citet{storrie92} performed one of the earliest attempts at morphological classification of galaxies using neural networks. |
Their data set consisted of 13 input features derived from images of galaxies which were then used to classily the galaxies into five classes: E. 80. 5a — Sb. Sc + Sd. and Ir. | Their data set consisted of 13 input features derived from images of galaxies which were then used to classify the galaxies into five classes: E, S0, Sa + Sb, Sc + Sd, and Irr. |
We used (heir input data set of 5217 galaxies. | We used their input data set of 5217 galaxies. |
The features in this data set are described in Storrie-Lombarclietal. | The features in this data set are described in \citet{storrie92}. |
(1992).. Dazell&Aha(2001) describes the use of this data set for galaxy classification using ensembles of neural networks. | \citet{bazell01} describes the use of this data set for galaxy classification using ensembles of neural networks. |
For our studies we eliminated one of the features. Ef. which is the error in an ellipse fit to B isophotes. | For our studies we eliminated one of the features, $E^{Fit}_{Err}$, which is the error in an ellipse fit to B isophotes. |
This feature had very small variance and equaled zero for approximately of the objects. | This feature had very small variance and equaled zero for approximately of the objects. |
Thus. we used 12 of the 13 features in the original data set. | Thus, we used 12 of the 13 features in the original data set. |
We also used a data set with an order of magnitude more objects than the ESOLV data. | We also used a data set with an order of magnitude more objects than the ESOLV data. |
The SDSS data consists of 54007 objects drawn from seven different classes. | The SDSS data consists of 54007 objects drawn from seven different classes. |
Each object is described by a total of six features: photometric values in u. g. r. 1. and z. and the reclshilt ol each object. | Each object is described by a total of six features: photometric values in u, g, r, i, and z, and the redshift of each object. |
Tables 1 and 2. summarize (he properties of the data sets we used. | Tables \ref{tab:esolv} and \ref{tab:sdss} summarize the properties of the data sets we used. |
For each class the tables show the munber of objects in the class. the percentage of total objects that. class | For each class the tables show the number of objects in the class, the percentage of total objects that class |
Let us asstune that the object consists of 7 poiut sources with intensities Aj aud positious rj. dj relative the reference poiut (j=l...n). | Let us assume that the object consists of $n$ point sources with intensities $A_j$ and positions $x_j$, $y_j$ relative the reference point $j=1 \dots n$ ). |
In general Aj. ry. yy vary with time. and so may be different for the different transits of the same object. | In general $A_j$, $x_j$, $y_j$ vary with time, and so may be different for the different transits of the same object. |
Given a specific model of the ob,ect we express Aj. irj; yj) as fuuctious of time f aud a set of nodel parameters α. | Given a specific model of the object we express $A_j$, $x_j$, $y_j$ as functions of time $t$ and a set of model parameters $\vec{a}$. |
For instance. in the case of a nou-variable orbital binary. à would consist of 15 parameuS. tthe five astrometric parameters of the mass centre. 1 magnitude of cach component. the mass ratio. ane seven elements for the relative orbit. | For instance, in the case of a non-variable orbital binary, $\vec{a}$ would consist of 15 parameters, the five astrometric parameters of the mass centre, the magnitude of each component, the mass ratio, and seven elements for the relative orbit. |
Generally speakiug. 16 object model is thus completely specified by à» and jo functions AQ.a). cj(f.a). yj(t.a@) for j=l...n. | Generally speaking, the object model is thus completely specified by $n$ and the functions $A_j(t,\vec{a})$, $x_j(t,\vec{a})$, $y_j(t,\vec{a})$ for $j=1 \dots n$. |
In 1e equations below we suppress. for brevity. the explicit lepeudeuce on ft aud a. | In the equations below we suppress, for brevity, the explicit dependence on $t$ and $\vec{a}$. |
For a given trausit the expected signal is moceled as je stun of the signals from the iudividual compoucuts. using Eqs. C12) | For a given transit the expected signal is modeled as the sum of the signals from the individual components, using Eqs. \ref{eq:ikphi}) ) |
aud (5)). | and \ref{eq:phi}) ). |
Thus. Expanding the trigonometric functions and equating the terms with those iu Eq. (1)) | Thus, Expanding the trigonometric functions and equating the terms with those in Eq. \ref{eq:ik}) ) |
vields Recall that Aj. ej. yj) depend ou the model parauicters a. | yields Recall that $A_j$, $x_j$, $y_j$ depend on the model parameters $\vec{a}$. |
The general procedure is then to adjust @ iu such a wav that. for the whole set. of transits. the calculates signal parameters 54 b; from Eq. (11)) | The general procedure is then to adjust $\vec{a}$ in such a way that, for the whole set of transits, the calculated signal parameters $b_1$ $b_5$ from Eq. \ref{eq:phaseelem}) ) |
agree. as well as yossible. with the observed values. | agree, as well as possible, with the observed values. |
The ajustiueut nav use the weielted least-squares method. using the staudarc errors of the observed signal parameters to set the weights: suit other (aud more robust) metrics cau also be used. | The adjustment may use the weighted least-squares method, using the standard errors of the observed signal parameters to set the weights; but other (and more robust) metrics can also be used. |
Iu oeeneral the problem can be formulated as a constraiuec uinnuuzatiou prodem in the mnultidiueusional iode aralneter space. | In general the problem can be formulated as a constrained minimization problem in the multi-dimensional model parameter space. |
The trigonometric functious iu Eq. (11)) | The trigonometric functions in Eq. \ref{eq:phaseelem}) ) |
mean that he signal parameters b; depend in a highlv non-liucar nanner ou the model paramcters which affect 6; aud gj. | mean that the signal parameters $b_i$ depend in a highly non-linear manner on the model parameters which affect $x_j$ and $y_j$. |
For instance. im terms of a displacement ofone of the point sources. the effect on b, aud b; is approximately linear ouly for displacements less than about 1/2f20.1 aresec. corresponding to 1 rad change in the modulation phase. | For instance, in terms of a displacement of one of the point sources, the effect on $b_4$ and $b_5$ is approximately linear only for displacements less than about $1/2f \simeq 0.1$ arcsec, corresponding to 1 rad change in the modulation phase. |
Iu the aperture svuthesis nuaenme this uou-linearitv is manifest in the complex. structure of the ‘dirty beau (Fie. £9) | In the aperture synthesis imaging this non-linearity is manifest in the complex structure of the `dirty beam' (Fig. \ref{fig:dbeam}) ) |
at all spatial scales larger than about 0.1 arcsec. | at all spatial scales larger than about 0.1 arcsec. |
Additional nou-lnearitices iu the complete object model may result from the geometrical description of the source positions. lin terms of orbital elements. | Additional non-linearities in the complete object model may result from the geometrical description of the source positions, in terms of orbital elements. |
The nou-luearity of the object model has two iuportaut consequences for the model fitting. | The non-linearity of the object model has two important consequences for the model fitting. |
Firstly. it is usually necessary to use a non-linear. iterative adjustinent algorithuu. such as the LevenbereMarquardt method (Press et al. 1992)). | Firstly, it is usually necessary to use a non-linear, iterative adjustment algorithm, such as the Levenberg–Marquardt method (Press et al. \cite{nr2}) ). |
Secondly. a good initial euess of the model paraiucters is usually required. | Secondly, a good initial guess of the model parameters is usually required. |
In particular the parameters directly affecting the positions of the »oiut sources need to be specified to withiu (vehat corresponds to) a few teuths of an aresec. | In particular the parameters directly affecting the positions of the point sources need to be specified to within (what corresponds to) a few tenths of an arcsec. |
Without a eood initiala guess. the adjustinent aleorithiu is likely to convereeex"p on sole local müninmun. typically resulting iu positional errors of (approximately) an integer number of exid periods. | Without a good initial guess, the adjustment algorithm is likely to converge on some local minimum, typically resulting in positional errors of (approximately) an integer number of grid periods. |
The correct solution. corresponding to the elobal minumiuni may in principle always be found through sufBcieutlv extensive searching of the parameter space. | The correct solution, corresponding to the global minimum, may in principle always be found through sufficiently extensive searching of the parameter space. |
Alternatively. sufficieutly good initial euesses of tle point source positions cau often be obtained from the aperture svuthesis imagine. | Alternatively, sufficiently good initial guesses of the point source positions can often be obtained from the aperture synthesis imaging. |
Various least-squares inodoel fitting procedures were used for the reduction of double aud multiple stars during he construction of the Iipparcos Catalogue (see \lignard et citemieu ando references therein) | Various least-squares model fitting procedures were used for the reduction of double and multiple stars during the construction of the Hipparcos Catalogue (see Mignard et \\cite{mign} and references therein). |
The double-star srocessing of the NDAC data. reduction ο (Sódderlijelii et citess92)}) essentially used the technique outlined. above. aking the so-called Case ITistory Files (a precursor to the TD) as input. | The double-star processing of the NDAC data reduction consortium (Södderhjelm et \\cite{ss92}) ) essentially used the technique outlined above, taking the so-called Case History Files (a precursor to the TD) as input. |
Perhaps the ereatest potential of the TD lies in he possibility to combine the DUipparcos data wit[um independeut observations frou other iustrunments and epochs. | Perhaps the greatest potential of the TD lies in the possibility to combine the Hipparcos data with independent observations from other instruments and epochs. |
For instance. full detezuination of a binary orbit eonerally requires data covering at least a whole period. | For instance, full determination of a binary orbit generally requires data covering at least a whole period. |
Ground-based speckle observations can somoetinies provide this. constraining the ecometry of the relative orbit iuc[um better than the Ilipparcos data alone. aud im turn leading to a better-deteriined space parallax. | Ground-based speckle observations can sometimes provide this, constraining the geometry of the relative orbit much better than the Hipparcos data alone, and in turn leading to a better-determined space parallax. |
Iu some favourable cases the location of the mass ceutre in the relative orbit) (aud hence the mass ratio) can be determined (Sodderhjchliu ct citess97:: SÓdderhjeha 1999)). | In some favourable cases the location of the mass centre in the relative orbit (and hence the mass ratio) can be determined (Södderhjelm et \\cite{ss97}; Södderhjelm \cite{ss99}) ). |
One coniplicatiou of the Ilpparcos double star processing has been the wide variety of applicable object models. and the consequent need to experiment and interact with the solutions. | One complication of the Hipparcos double star processing has been the wide variety of applicable object models, and the consequent need to experiment and interact with the solutions. |
This process may be much facilitated by usingo Ὁgeneral aud flexible software for the model fitting. rather than hiehly specialized routines, | This process may be much facilitated by using general and flexible software for the model fitting, rather than highly specialized routines. |
Au example of this is given below. | An example of this is given below. |
atmospheric opacity across the dillerent wavebands. whereas we are comparing our data to WVM cata. | atmospheric opacity across the different wavebands, whereas we are comparing our data to WVM data. |
Nevertheless. recent work shows that provided all relevant. effects: are taken into consideration. the results from LTS ancl WVM measurements eenerallv agree. well (Pardo et al. | Nevertheless, recent work shows that provided all relevant effects are taken into consideration, the results from FTS and WVM measurements generally agree well (Pardo et al. |
2004). | 2004). |
llence our result. is fully consistent with previous work. | Hence our result is fully consistent with previous work. |
Therefore we can use this relation to calibrate our subsequent measurements. | Therefore we can use this relation to calibrate our subsequent measurements. |
The first astronomical object we imaged was Jupiter. | The first astronomical object we imaged was Jupiter. |
Figure 7 shows our ὀθθ-μαι. map of Jupiter. | Figure \ref{jupiter} shows our $\mu$ m map of Jupiter. |
I was constructed. from two consecutive maps of Jupiter. taken immeclately one after the other. over the airmass range 1.25 to 1.27. | It was constructed from two consecutive maps of Jupiter, taken immediately one after the other, over the airmass range 1.25 to 1.27. |
Phe two maps both show the same structure. and so they were co-adcded to increase the signal-to-noise ratio. | The two maps both show the same structure, and so they were co-added to increase the signal-to-noise ratio. |
We note that the source is not exactly centred in the image. being misaligned by ~30 aresec. | We note that the source is not exactly centred in the image, being misaligned by $\sim$ 30 arcsec. |
“Phe telescope absolute pointing accuracy (based on measurements made with other instruments) at this time was : <3 arcsec. | The telescope absolute pointing accuracy (based on measurements made with other instruments) at this time was $\leq$ 3 arcsec. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.