Python numpy sampling a 2D array for n rows-CodePudding

I have a numpy array as follows, I want to take a random sample of n rows.

             ([[996.924, 265.879, 191.655],
             [996.924, 265.874, 191.655],
             [996.925, 265.884, 191.655],
             [997.294, 265.621, 192.224],
             [997.294, 265.643, 192.225],
             [997.304, 265.652, 192.223]], dtype=float32)

I've tried:

rows_id = random.sample(range(0,arr.shape[1]-1), 1)
row = arr[rows_id, :]

But this 9ndex mask only returns a single row, I want to return n rows as an numpy array (without duplication).

CodePudding user response：

You have three key issues: arr.shape[1] returns the number of columns, while you want the number of rows--arr.shape[0]. Second, the second parameter to range is exclusive, so you don't really need the -1. Third, your last parameter to random.sample is the number of rows, which you set to 1.

A better way to do what you're trying might be random.choices instead.

CodePudding user response：

Try where x is your original array:

n = 2  #number of rows
idx = np.random.choice(len(x), n, replace = False)
result = np.array([x[i] for i in idx])