如何按降序对二维数组中的一半进行排序（numpy）

Question

I'm trying to create an array (10000, 50) size (I'm mentioning the size because efficiency is important), and then :

• Sort the first 5000 rows in ascending order
• Sort the next 5000 rows in descending order.

Here is my code :

samples = 10  # I'm going to increase it 10000
sampleLength = 4 # I'm going to increase it 50
halfSamples = int(samples/2)

xx = numpy.multiply(10, numpy.random.random((samples, sampleLength)))
xx[0:halfSamples,0:sampleLength]=numpy.sort(xx[0:halfSamples,0:sampleLength],axis=1)
xx[halfSamples:samples,0:sampleLength]=numpy.sort(xx[halfSamples:samples,0:sampleLength],axis=1)

This sorts both half of the array in ascending order, the only thing I can't find is what parameter to give in my last line to make it in a descending order.

I've tried based on this link : Reverse sort a 2d numpy array in python

xx[halfSamples:samples,0:sampleLength]=numpy.sort(xx[halfSamples:samples,0:sampleLength:-1],axis=1)

But got an error :

ValueError: could not broadcast input array from shape (5,0) into shape (5,4)

Thanks

Answer 1

It would probably be faster to sort the array in place using its .sort method, rather than np.sort which returns a copy. You can index the second dimension using a negative step size to sort the columns of the last 5000 rows in descending order:

x = np.random.randn(10000, 50)
x[:5000].sort(axis=1)
x[-5000:, ::-1].sort(axis=1)