All Figures

$\begin{figure} \par\includegraphics[angle=360,width=7cm,clip]{7638f1.eps} \end{figure}$ Figure 1: Schematic overview of the different steps (sections indicated) in the development and comparison of the classification methods presented in this paper.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[angle=180,width=14.5cm,clip]{7638f2.ps} \end{figure}$ Figure 2: The range for the frequency f₁ (in cycles/day), its first harmonic amplitude A₁₁ (in magnitude), the phases PH₁₂ (in radians) and the variance ratio v_f1/v (varrat) for all the 35 considered variability classes listed in Table 1. For visibility reasons, we have plotted the logarithm of the frequency and amplitude values. Every symbol in the plots corresponds to the parameter value of exactly one light curve. In this way, we attempt to visualize the distribution of the light curve parameters, in addition to their mere range.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[angle=-90,width=14.2cm,clip]{7638f3.eps} \end{figure}$ Figure 3: Box-and-whiskers plot of the logarithm of f₁ for 29 classes with sufficient members to define such tools in the training set. Central boxes represent the median and interquantile ranges (25 to 75%) and the outer whiskers represent rule-of-thumb boundaries for the definition of outliers (1.5 the quartile range). The box widths are proportional to the number of examples in the class.
Open with DEXTER

In the text

$\begin{figure} \includegraphics[origin=rb,angle=-90,width=12cm,clip]{7638f4.eps} \end{figure}$ Figure 4: Hexagonal representation of the two dimensional density of examples of the Classical Cepheids class in the $\log(P)$ - $\log(R_{21})$ space. The quantity $R_{21}\equiv A_{12}/A_{11}$ represents the ratio of the second to the first harmonic amplitude of f₁. This plot is comparable to Fig. 3 of Udalski et al. (1999b).
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=15cm,angle=180,scale=0.94,clip]{7638f5.ps} \end{figure}$ Figure 5: The range in amplitudes A_1j for the 3 higher harmonics of f₁, and the linear trend b. For visibility reasons, we have plotted the logarithm of the amplitude values.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=15cm,angle=180,scale=0.94,clip]{7638f6.ps} \end{figure}$ Figure 6: The range for the frequencies f₂ and f₃ and the phases PH_1j of the higher harmonics of f₁. For visibility reasons, we have plotted the logarithm of the frequency values. Note the split into two clouds of the phase values PH₁₃ for the eclipsing binary classes. This is a computational artefact: phase values close to $-\pi$ are equivalent to values close to $+\pi$ , so the clouds actually represent a single cloud.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=15cm,angle=180,scale=0.94,clip]{7638f7.ps} \end{figure}$ Figure 7: The range in amplitudes A_2j for the 4 harmonics of f₂. For visibility reasons, we have plotted the logarithm of the amplitude values.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=15cm,angle=180,scale=0.94,clip]{7638f8.ps} \end{figure}$ Figure 8: The range in amplitudes A_3j for the 4 harmonics of f₃. For visibility reasons, we have plotted the logarithm of the amplitude values.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=15cm,angle=180,scale=0.94,clip]{7638f9.ps} \end{figure}$ Figure 9: The range in phases PH_2j for the 4 harmonics of f₂. As can be seen from the plots, the distribution of these parameters is rather uniform for every class. They are unlikely to be good classification parameters, since for none of the classes, clear clustering of the phase values is present.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[width=15cm,angle=180,scale=0.94,clip]{7638f10.ps} \end{figure}$ Figure 10: The range in phases PH_3j for the 4 harmonics of f₃. The same comments as for Fig. 9 apply here.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[angle=-90,width=13.8cm,clip]{7638f11.eps} \end{figure}$ Figure 11: Box-and-whiskers plot of the logarithm of R₂₁ for all classes in the training set. Central boxes represent the median and interquantile ranges (25 to 75%) and the outer whiskers represent rule-of-thumb boundaries for the definition of outliers (1.5 the quartile range). The boxes widths are proportional to the number of examples in the class.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[angle=-90,width=13.8cm,clip]{7638f12.eps} \end{figure}$ Figure 12: Box-and-whiskers plot of the logarithm of the variance ratio v_f1/v (varrat) for all classes in the training set. Central boxes represent the median and interquantile ranges (25 to 75%) and the outer whiskers represent rule-of-thumb boundaries for the definition of outliers (1.5 the quartile range). The boxes widths are proportional to the number of examples in the class.
Open with DEXTER

In the text

$\begin{figure} \par\includegraphics[angle=360,width=7cm,clip]{7638f1.eps} \end{figure}$	Figure 1: Schematic overview of the different steps (sections indicated) in the development and comparison of the classification methods presented in this paper.
Open with DEXTER

$\begin{figure} \includegraphics[origin=rb,angle=-90,width=12cm,clip]{7638f4.eps} \end{figure}$	Figure 4: Hexagonal representation of the two dimensional density of examples of the Classical Cepheids class in the $\log(P)$ - $\log(R_{21})$ space. The quantity $R_{21}\equiv A_{12}/A_{11}$ represents the ratio of the second to the first harmonic amplitude of f₁. This plot is comparable to Fig. 3 of Udalski et al. (1999b).
Open with DEXTER

$\begin{figure} \par\includegraphics[width=15cm,angle=180,scale=0.94,clip]{7638f5.ps} \end{figure}$	Figure 5: The range in amplitudes A_1j for the 3 higher harmonics of f₁, and the linear trend b. For visibility reasons, we have plotted the logarithm of the amplitude values.
Open with DEXTER

$\begin{figure} \par\includegraphics[width=15cm,angle=180,scale=0.94,clip]{7638f7.ps} \end{figure}$	Figure 7: The range in amplitudes A_2j for the 4 harmonics of f₂. For visibility reasons, we have plotted the logarithm of the amplitude values.
Open with DEXTER

$\begin{figure} \par\includegraphics[width=15cm,angle=180,scale=0.94,clip]{7638f8.ps} \end{figure}$	Figure 8: The range in amplitudes A_3j for the 4 harmonics of f₃. For visibility reasons, we have plotted the logarithm of the amplitude values.
Open with DEXTER

$\begin{figure} \par\includegraphics[width=15cm,angle=180,scale=0.94,clip]{7638f9.ps} \end{figure}$	Figure 9: The range in phases PH_2j for the 4 harmonics of f₂. As can be seen from the plots, the distribution of these parameters is rather uniform for every class. They are unlikely to be good classification parameters, since for none of the classes, clear clustering of the phase values is present.
Open with DEXTER

$\begin{figure} \par\includegraphics[width=15cm,angle=180,scale=0.94,clip]{7638f10.ps} \end{figure}$	Figure 10: The range in phases PH_3j for the 4 harmonics of f₃. The same comments as for Fig. 9 apply here.
Open with DEXTER

$\begin{figure} \par\includegraphics[angle=-90,width=13.8cm,clip]{7638f11.eps} \end{figure}$	Figure 11: Box-and-whiskers plot of the logarithm of R₂₁ for all classes in the training set. Central boxes represent the median and interquantile ranges (25 to 75%) and the outer whiskers represent rule-of-thumb boundaries for the definition of outliers (1.5 the quartile range). The boxes widths are proportional to the number of examples in the class.
Open with DEXTER

$\begin{figure} \par\includegraphics[angle=-90,width=13.8cm,clip]{7638f12.eps} \end{figure}$	Figure 12: Box-and-whiskers plot of the logarithm of the variance ratio v_f1/v (varrat) for all classes in the training set. Central boxes represent the median and interquantile ranges (25 to 75%) and the outer whiskers represent rule-of-thumb boundaries for the definition of outliers (1.5 the quartile range). The boxes widths are proportional to the number of examples in the class.
Open with DEXTER