减去两个不同形状的 np.ndarray

Question

ndarray of different shapes and the bigger one has a shape of (22470,2) and it looks like this

df1

array([[-0.39911392,  0.46759156],
       [ 0.28343494,  0.88479157],
       [-0.0114085 , -1.23768313],
       ...,
       [-0.35930586,  0.54784439],
       [-0.37994004,  0.51332771],
       [-0.36309593,  0.49318486]])

and the small one which represents the outliers of df1 array and its shape is (675,2) and it looks like this

df2
array([[-0.04450032,  0.31053589],
       [-0.4320086 ,  0.14815988],
       [-0.07948631, -1.32638555],
       ...,
       [-0.32619787,  0.34910699],
       [-0.50870225, -0.230849  ],
       [-0.43532727,  0.49763502]])

so tried to subtract both of them to have a new array that contains everything in df1 except df2 but it gives me this error

ValueError: operands could not be broadcast together with shapes (22470,2) (675,2)

How can I do it in Python?

Answer 1

"Subtracting" two arrays does not perform set operations on the arrays, it simply subtracts the values of one from the values of the other (ie 4 - 3 => 1).

What you want to do is basically a set operation. There is no simple straightforward way to do what you want, how you have presented it (but that doesn't mean it can't be done). Comparing floating-point numbers for exact equality is a bad idea, instead you will find it much more useful to collect an array of indices of the outliers rather than their values. Then you can index your array like this question .

So this would then be something like

df1 = array([[1.234, 2.345], [3.3452, 2.456], [5.234, 7.453]])

# This is an array of indices, not float values.
df2 = array([1])

keep = np.ones(len(df1), dtype=bool)
keep[df2] = 0
newdf = df1[keep]

# newdf: [[1.234, 2.345], [5.234, 7.453]]