Coarsen array by applying reduction to fixed size neighborhoods Parameters ---------- reduction: function Function like np.sum, np.mean, etc... x: np.ndarray Array to be coarsened axes: dict Mapping of axis to coarsening factor Examples --------
(reduction, x, axes, trim_excess=False, **kwargs)
| 83 | |
| 84 | |
| 85 | def coarsen(reduction, x, axes, trim_excess=False, **kwargs): |
| 86 | """Coarsen array by applying reduction to fixed size neighborhoods |
| 87 | |
| 88 | Parameters |
| 89 | ---------- |
| 90 | reduction: function |
| 91 | Function like np.sum, np.mean, etc... |
| 92 | x: np.ndarray |
| 93 | Array to be coarsened |
| 94 | axes: dict |
| 95 | Mapping of axis to coarsening factor |
| 96 | |
| 97 | Examples |
| 98 | -------- |
| 99 | >>> x = np.array([1, 2, 3, 4, 5, 6]) |
| 100 | >>> coarsen(np.sum, x, {0: 2}) |
| 101 | array([ 3, 7, 11]) |
| 102 | >>> coarsen(np.max, x, {0: 3}) |
| 103 | array([3, 6]) |
| 104 | |
| 105 | Provide dictionary of scale per dimension |
| 106 | |
| 107 | >>> x = np.arange(24).reshape((4, 6)) |
| 108 | >>> x |
| 109 | array([[ 0, 1, 2, 3, 4, 5], |
| 110 | [ 6, 7, 8, 9, 10, 11], |
| 111 | [12, 13, 14, 15, 16, 17], |
| 112 | [18, 19, 20, 21, 22, 23]]) |
| 113 | |
| 114 | >>> coarsen(np.min, x, {0: 2, 1: 3}) |
| 115 | array([[ 0, 3], |
| 116 | [12, 15]]) |
| 117 | |
| 118 | You must avoid excess elements explicitly |
| 119 | |
| 120 | >>> x = np.array([1, 2, 3, 4, 5, 6, 7, 8]) |
| 121 | >>> coarsen(np.min, x, {0: 3}, trim_excess=True) |
| 122 | array([1, 4]) |
| 123 | """ |
| 124 | # Insert singleton dimensions if they don't exist already |
| 125 | for i in range(x.ndim): |
| 126 | if i not in axes: |
| 127 | axes[i] = 1 |
| 128 | |
| 129 | if trim_excess: |
| 130 | ind = tuple( |
| 131 | slice(0, -(d % axes[i])) if d % axes[i] else slice(None, None) |
| 132 | for i, d in enumerate(x.shape) |
| 133 | ) |
| 134 | x = x[ind] |
| 135 | |
| 136 | # (10, 10) -> (5, 2, 5, 2) |
| 137 | newshape = tuple(concat([(x.shape[i] // axes[i], axes[i]) for i in range(x.ndim)])) |
| 138 | |
| 139 | return reduction(x.reshape(newshape), axis=tuple(range(1, x.ndim * 2, 2)), **kwargs) |
| 140 | |
| 141 | |
| 142 | def trim(x, axes=None): |