时间戳

首页 > 解决方案 > Xarray：使同一Dataset中的两个DataArrays使用相同的坐标系

python - Xarray：使同一Dataset中的两个DataArrays使用相同的坐标系

问题描述

我有一个 ArviZ InferenceData 后验轨迹，它是一个 XArray 数据集。

在那里，我的两个随机变量的后验轨迹a_mu_org是b_mu_orgDataArrays。它们的坐标是：

从语义上讲，a_mu_org实际上b_mu_org应该由 15 个生物的单一分类坐标系索引，而不是单独的索引。

为了更清楚一点，这里是完整的数据集字符串 repr：

<xarray.Dataset>
Dimensions:             (L_dim_0: 34281, a_dim_0: 456260, a_prot_shift_dim_0: 34281, b_dim_0: 456260, b_mu_org_dim_0: 15, b_prot_shift_dim_0: 34281, chain: 1, draw: 2000, organism: 15, sigma_dim_0: 34281, t50_org_dim_0: 15, t50_prot_dim_0: 39957)
Coordinates:
  * chain               (chain) int64 0
  * draw                (draw) int64 0 1 2 3 4 5 ... 1995 1996 1997 1998 1999
  * a_prot_shift_dim_0  (a_prot_shift_dim_0) object 'A0A023PXQ4_YMR173W-A' ... 'Z4YNA9_AB124611'
  * b_prot_shift_dim_0  (b_prot_shift_dim_0) object 'A0A023PXQ4_YMR173W-A' ... 'Z4YNA9_AB124611'
  * L_dim_0             (L_dim_0) object 'A0A023PXQ4_YMR173W-A' ... 'Z4YNA9_AB124611'
    a_mu_org_dim_0      (organism) int64 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
  * a_dim_0             (a_dim_0) object 'ytzI' 'mtlF' ... 'atpG2' 'atpB2'
  * b_mu_org_dim_0      (b_mu_org_dim_0) int64 0 1 2 3 4 5 ... 9 10 11 12 13 14
  * b_dim_0             (b_dim_0) object 'ytzI' 'mtlF' ... 'atpG2' 'atpB2'
  * t50_prot_dim_0      (t50_prot_dim_0) <U65 'Bacillus subtilis_168_lysate_R1-C0H3Q1_ytzI' ... 'Oleispira antarctica_RB-8_lysate_R1-R4YVF0_atpB2'
  * t50_org_dim_0       (t50_org_dim_0) <U43 'Arabidopsis thaliana seedling lysate' ... 'Thermus thermophilus HB27 lysate'
  * sigma_dim_0         (sigma_dim_0) object 'A0A023PXQ4_YMR173W-A' ... 'Z4YNA9_AB124611'
Dimensions without coordinates: organism
Data variables:
    a_org_pop           (chain, draw) float32 519.3236 518.8292 ... 517.84784
    a_prot_shift        (chain, draw, a_prot_shift_dim_0) float32 ...
    b_org_pop           (chain, draw) float32 11.509291 11.445394 ... 11.929538
    b_prot_shift        (chain, draw, b_prot_shift_dim_0) float32 ...
    L_pop               (chain, draw) float32 3.445896 3.4300675 ... 3.3917112
    L                   (chain, draw, L_dim_0) float32 ...
    a_mu_org            (chain, draw, organism) float32 430.56827 ... 813.2518
    a                   (chain, draw, a_dim_0) float32 ...
    b_mu_org            (chain, draw, b_mu_org_dim_0) float32 9.997488 ... 8.389757
    b                   (chain, draw, b_dim_0) float32 ...
    t50_prot            (chain, draw, t50_prot_dim_0) float32 39.249863 ... 52.19809
    t50_org             (chain, draw, t50_org_dim_0) float32 43.067646 ... 96.93388
    sigma               (chain, draw, sigma_dim_0) float32 ...
Attributes:
    created_at:                 2020-04-23T08:54:58.300091
    arviz_version:              0.7.0
    inference_library:          pymc3
    inference_library_version:  3.8

我想制作a_mu_org并b_mu_org采用尺寸 ( chain, draw, organism) 而不是它们单独的a_mu_organd b_mu_org。我已经尝试过的事情包括：

我想要完成的事情可能吗？

标签： pythonpython-xarray

解决方案

我不确定我的解决方案是不是很好的做法，感觉有点太老套了。此外，术语非常棘手，我将尝试坚持使用xarray 术语，但这样做可能会失败。诀窍是删除坐标a_dim_0，使其b_dim_0成为唯一尺寸（现在没有坐标的尺寸）。之后，可以将它们重命名为相同的事物并分配给新的坐标。这是一个例子：

从以下名为的数据集开始ds：

<xarray.Dataset>
Dimensions:  (a_dim_0: 15, b_dim_0: 15, chain: 4, draw: 100)
Coordinates:
  * chain    (chain) int64 0 1 2 3
  * draw     (draw) int64 0 1 2 3 4 5 6 7 8 9 ... 90 91 92 93 94 95 96 97 98 99
  * a_dim_0  (a_dim_0) int64 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
  * b_dim_0  (b_dim_0) int64 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Data variables:
    a        (chain, draw, a_dim_0) float64 0.8152 1.189 ... 1.32 -0.2023
    b        (chain, draw, b_dim_0) float64 0.6447 -0.8059 ... -0.06435 -0.8666

以下 3 个命令可以解决问题（ 的位置assign_coord似乎不会影响输出，这是有道理的，但首先删除坐标然后重命名是关键）：

organism_names = [f"o{i}" for i in range(15)]
ds.reset_index(["a_dim_0", "b_dim_0"], drop=True) \
    .assign_coords(organism=organism_names) \
    .rename({"a_dim_0": "organism", "b_dim_0": "organism"})

输出：

<xarray.Dataset>
Dimensions:   (chain: 4, draw: 100, organism: 15)
Coordinates:
  * chain     (chain) int64 0 1 2 3
  * draw      (draw) int64 0 1 2 3 4 5 6 7 8 9 ... 90 91 92 93 94 95 96 97 98 99
  * organism  (organism) <U3 'o0' 'o1' 'o2' 'o3' ... 'o11' 'o12' 'o13' 'o14'
Data variables:
    a         (chain, draw, organism) float64 0.8152 1.189 ... 1.32 -0.2023
    b         (chain, draw, organism) float64 0.6447 -0.8059 ... -0.8666

推荐阅读