|
1 | | -import numpy as np |
2 | | -from typing import Tuple |
3 | | - |
4 | | -from autoconf import cached_property |
5 | | - |
6 | | -import autoarray as aa |
7 | | - |
8 | | - |
9 | | -class DatasetInterp: |
10 | | - def __init__(self, dataset: aa.Imaging): |
11 | | - """ |
12 | | - An ellipse interpolator, which contains a dataset (e.g. the image data and noise-map) and performs interpo.aiton |
13 | | - calculations used for ellipse fitting. |
14 | | -
|
15 | | - This object is used by the input to the `FitEllipse` object, which fits the dataset with ellipses and quantifies |
16 | | - the goodness-of-fit via a residual map, likelihood, chi-squared and other quantities. |
17 | | -
|
18 | | - The following quantities of the ellipse data are interpolated and used for the following tasks: |
19 | | -
|
20 | | - - `data`: The image data, which shows the signal that is analysed and fitted with ellipses. |
21 | | -
|
22 | | - - `noise_map`: The RMS standard deviation error in every pixel, which is used to compute the chi-squared value |
23 | | - and likelihood of a fit. |
24 | | -
|
25 | | - The `data` and `noise_map` are typically the same images of a galaxy used to perform standard light-profile |
26 | | - fitting. |
27 | | -
|
28 | | - Parameters |
29 | | - ---------- |
30 | | - dataset |
31 | | - The imaging data, containing the image data, noise map. |
32 | | - """ |
33 | | - self.dataset = dataset |
34 | | - |
35 | | - @cached_property |
36 | | - def points_interp(self) -> Tuple[np.ndarray, np.ndarray]: |
37 | | - """ |
38 | | - The points on which the interpolation from the 2D grid of data is performed. |
39 | | - """ |
40 | | - |
41 | | - x = self.dataset.mask.derive_grid.all_false.native[0, :, 1] |
42 | | - y = np.flip(self.dataset.mask.derive_grid.all_false.native[:, 0, 0]) |
43 | | - |
44 | | - return (x, y) |
45 | | - |
46 | | - @cached_property |
47 | | - def mask_interp(self) -> "interpolate.RegularGridInterpolator": |
48 | | - """ |
49 | | - Returns a 2D interpolation of the mask, which is used to determine whether inteprolated values use a masked |
50 | | - pixel for the interpolation and thus should not be included in a fit. |
51 | | - """ |
52 | | - from scipy import interpolate |
53 | | - |
54 | | - return interpolate.RegularGridInterpolator( |
55 | | - points=self.points_interp, |
56 | | - values=np.float64(self.dataset.data.mask), |
57 | | - bounds_error=False, |
58 | | - fill_value=0.0, |
59 | | - ) |
60 | | - |
61 | | - @cached_property |
62 | | - def data_interp(self) -> "interpolate.RegularGridInterpolator": |
63 | | - """ |
64 | | - Returns a 2D interpolation of the data, which is used to evaluate the data at any point in 2D space. |
65 | | - """ |
66 | | - from scipy import interpolate |
67 | | - |
68 | | - return interpolate.RegularGridInterpolator( |
69 | | - points=self.points_interp, |
70 | | - values=np.float64(self.dataset.data.native), |
71 | | - bounds_error=False, |
72 | | - fill_value=0.0, |
73 | | - ) |
74 | | - |
75 | | - @cached_property |
76 | | - def noise_map_interp(self) -> "interpolate.RegularGridInterpolator": |
77 | | - """ |
78 | | - Returns a 2D interpolation of the noise-map, which is used to evaluate the noise-map at any point in 2D space. |
79 | | - """ |
80 | | - from scipy import interpolate |
81 | | - |
82 | | - return interpolate.RegularGridInterpolator( |
83 | | - points=self.points_interp, |
84 | | - values=np.float64(self.dataset.noise_map.native), |
85 | | - bounds_error=False, |
86 | | - fill_value=0.0, |
87 | | - ) |
| 1 | +import numpy as np |
| 2 | +from typing import Tuple |
| 3 | + |
| 4 | +from autoconf import cached_property |
| 5 | + |
| 6 | +import autoarray as aa |
| 7 | + |
| 8 | + |
| 9 | +class DatasetInterp: |
| 10 | + def __init__(self, dataset: aa.Imaging): |
| 11 | + """ |
| 12 | + An ellipse interpolator, which contains a dataset (e.g. the image data and noise-map) and performs interpo.aiton |
| 13 | + calculations used for ellipse fitting. |
| 14 | +
|
| 15 | + This object is used by the input to the `FitEllipse` object, which fits the dataset with ellipses and quantifies |
| 16 | + the goodness-of-fit via a residual map, likelihood, chi-squared and other quantities. |
| 17 | +
|
| 18 | + The following quantities of the ellipse data are interpolated and used for the following tasks: |
| 19 | +
|
| 20 | + - `data`: The image data, which shows the signal that is analysed and fitted with ellipses. |
| 21 | +
|
| 22 | + - `noise_map`: The RMS standard deviation error in every pixel, which is used to compute the chi-squared value |
| 23 | + and likelihood of a fit. |
| 24 | +
|
| 25 | + The `data` and `noise_map` are typically the same images of a galaxy used to perform standard light-profile |
| 26 | + fitting. |
| 27 | +
|
| 28 | + Parameters |
| 29 | + ---------- |
| 30 | + dataset |
| 31 | + The imaging data, containing the image data, noise map. |
| 32 | + """ |
| 33 | + self.dataset = dataset |
| 34 | + |
| 35 | + @cached_property |
| 36 | + def points_interp(self) -> Tuple[np.ndarray, np.ndarray]: |
| 37 | + """ |
| 38 | + The points on which the interpolation from the 2D grid of data is performed. |
| 39 | + """ |
| 40 | + |
| 41 | + x = self.dataset.mask.derive_grid.all_false.native[0, :, 1] |
| 42 | + y = np.flip(self.dataset.mask.derive_grid.all_false.native[:, 0, 0]) |
| 43 | + |
| 44 | + return (x, y) |
| 45 | + |
| 46 | + def mask_interp(self, points, xp=np): |
| 47 | + """ |
| 48 | + Returns a 2D interpolation of the mask, which is used to determine whether inteprolated values use a masked |
| 49 | + pixel for the interpolation and thus should not be included in a fit. |
| 50 | +
|
| 51 | + Backwards-compatibility note: this used to be a cached property that |
| 52 | + returned a scipy ``RegularGridInterpolator`` instance, also callable on |
| 53 | + ``(N, 2)`` points. Existing call sites in ``fit_ellipse.py`` that read |
| 54 | + ``self.interp.mask_interp(points)`` continue to work; the sentinel |
| 55 | + check ``self.interp.mask_interp is not None`` also continues to pass |
| 56 | + (a bound method is never None). |
| 57 | +
|
| 58 | + Parameters |
| 59 | + ---------- |
| 60 | + points |
| 61 | + An ``(N, 2)`` array of query coordinates. Out-of-bounds queries |
| 62 | + return 0.0 (treat outside-the-grid as unmasked). |
| 63 | + xp |
| 64 | + Array namespace (``numpy`` or ``jax.numpy``). Defaults to |
| 65 | + ``numpy``. |
| 66 | + """ |
| 67 | + x_axis, y_axis = self.points_interp |
| 68 | + return aa.interp_2d( |
| 69 | + points, |
| 70 | + x_axis, |
| 71 | + y_axis, |
| 72 | + np.float64(self.dataset.data.mask), |
| 73 | + fill_value=0.0, |
| 74 | + xp=xp, |
| 75 | + ) |
| 76 | + |
| 77 | + def data_interp(self, points, xp=np): |
| 78 | + """ |
| 79 | + Returns a 2D interpolation of the data, which is used to evaluate the data at any point in 2D space. |
| 80 | +
|
| 81 | + Parameters |
| 82 | + ---------- |
| 83 | + points |
| 84 | + An ``(N, 2)`` array of query coordinates. Out-of-bounds queries |
| 85 | + return 0.0. |
| 86 | + xp |
| 87 | + Array namespace (``numpy`` or ``jax.numpy``). Defaults to |
| 88 | + ``numpy``. |
| 89 | + """ |
| 90 | + x_axis, y_axis = self.points_interp |
| 91 | + return aa.interp_2d( |
| 92 | + points, |
| 93 | + x_axis, |
| 94 | + y_axis, |
| 95 | + np.float64(self.dataset.data.native), |
| 96 | + fill_value=0.0, |
| 97 | + xp=xp, |
| 98 | + ) |
| 99 | + |
| 100 | + def noise_map_interp(self, points, xp=np): |
| 101 | + """ |
| 102 | + Returns a 2D interpolation of the noise-map, which is used to evaluate the noise-map at any point in 2D space. |
| 103 | +
|
| 104 | + Parameters |
| 105 | + ---------- |
| 106 | + points |
| 107 | + An ``(N, 2)`` array of query coordinates. Out-of-bounds queries |
| 108 | + return 0.0. |
| 109 | + xp |
| 110 | + Array namespace (``numpy`` or ``jax.numpy``). Defaults to |
| 111 | + ``numpy``. |
| 112 | + """ |
| 113 | + x_axis, y_axis = self.points_interp |
| 114 | + return aa.interp_2d( |
| 115 | + points, |
| 116 | + x_axis, |
| 117 | + y_axis, |
| 118 | + np.float64(self.dataset.noise_map.native), |
| 119 | + fill_value=0.0, |
| 120 | + xp=xp, |
| 121 | + ) |
0 commit comments