使用 TSNE 进行降维。为什么 3D 图形不起作用？

Question

我使用了来自 Sklearn 的 Digits 数据集，并尝试使用 TSNE(t-Distributed Stochastic Neighbor Embedding) 将维度从 64 减少到 3：

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
#%matplotib inline
from sklearn.manifold import TSNE
from sklearn.datasets import load_digits
from mpl_toolkits.mplot3d import Axes3D


digits = load_digits()
digits_df = pd.DataFrame(digits.data,)
digits_df["target"] = pd.Series(digits.target)

tsne = TSNE(n_components=3)
digits_tsne = tsne.fit_transform(digits_df.iloc[:,:64])
digits_df_tsne = pd.DataFrame(digits_tsne,
                            columns =["Component1","Component2","Component3"])

finalDf = pd.concat([digits_df_tsne, digits_df["target"]], axis = 1)

#Visualizing 3D
figure = plt.figure(figsize=(9,9))
axes = figure.add_subplot(111,projection = "3d")
dots = axes.scatter(xs = finalDf[:,0],ys = finalDf[:,1],zs = finalDf[:,2],
                   c = digits.target, cmap = plt.cm.get_cmap("nipy_spectral_r",10))

finalDf：

错误：

TypeError: '(slice(None, None, None), 0)' is an invalid key

怎么了？有人可以帮我吗？

Answer 1

您正在尝试对无效的 pandas 数据帧进行 numpy 切片，因此请先将数据帧转换为 numpy 数组。

这是更新后的代码：-

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
#%matplotib inline
from sklearn.manifold import TSNE
from sklearn.datasets import load_digits
from mpl_toolkits.mplot3d import Axes3D


digits = load_digits()
digits_df = pd.DataFrame(digits.data,)
digits_df["target"] = pd.Series(digits.target)

tsne = TSNE(n_components=3)
digits_tsne = tsne.fit_transform(digits_df.iloc[:,:64])
digits_df_tsne = pd.DataFrame(digits_tsne,
                            columns =["Component1","Component2","Component3"])

finalDf = pd.concat([digits_df_tsne, digits_df["target"]], axis = 1)

#Visualizing 3D
figure = plt.figure(figsize=(9,9))
axes = figure.add_subplot(111,projection = "3d")
dots = axes.scatter(xs = finalDf.to_numpy()[:,0],ys = finalDf.to_numpy()[:,1],zs = finalDf.to_numpy()[:,2],
                   c = digits.target, cmap = plt.cm.get_cmap("nipy_spectral_r",10))