python - 如何在anaconda中的win10上运行umap？

Question

我正在尝试从他们的文档中运行 UMAP 的示例：

import sklearn.datasets
import pandas as pd
import numpy as np
import umap

pendigits = sklearn.datasets.load_digits()
mapper = umap.UMAP().fit(pendigits.data)

import umap.plot
umap.plot.points(mapper)

但我面临这个错误ModuleNotFoundError: No module named 'umap.plot'：：

ModuleNotFoundError Traceback（最近一次调用最后一次）在 10 mapper = umap.UMAP().fit(pendigits.data) 11 ---> 12 import umap.plot 13 umap.plot.points(mapper)

ModuleNotFoundError：没有名为“umap.plot”的模块

---

import sys, platform, umap

print(umap.__version__)
print(platform.platform())
print(sys.version)
print(dir(umap))

会给：

0.3.10
Windows-10-10.x.xxx-SP0
3.7.6 | packaged by conda-forge | (default, Jan  7 2020, 10:48:41) [MSC v.xx 64 bit (AMD64)]
['UMAP', '__builtins__', '__cached__', '__doc__', '__file__', '__loader__', '__name__', '__package__', '__path__', '__spec__', '__version__', 'distances', 'nndescent', 'numba', 'pkg_resources', 'rp_tree', 'sparse', 'spectral', 'umap_', 'utils']

---

显然，作为一种解决方法，可以这样做，但我更喜欢享受丰富的 umap 绘图功能：

import umap
from sklearn.datasets import load_digits
import matplotlib.pyplot as plt
%matplotlib inline

digits = load_digits()
embedding = umap.UMAP().fit_transform(digits.data)
plt.plot(embedding[:,0],embedding[:,1], 'o')

标签： pythonpython-3.xmachine-learninganacondapypi