NumPy: creating a two-dimensional array

All elements of the matrix b will have the data type float (most likely float64), since one element is 1.5 of the type float.

float - a more "strong" type compared to int, i.e. a value of type int (meaning Numpy data types, not Python int with unlimited precision) can be converted to float without loss of precision (information), and not back:

In [56]: b
Out[56]:
array([[1.5, 2. , 3. ],
       [4. , 5. , 6. ]])

In [57]: b.dtype
Out[57]: dtype('float64')

Indexing in Numpy is much more "advanced" compared to standard Python indexing. lists:

In [58]: b[1,0]
Out[58]: 4.0

Here, before the comma, it is the index (s) of the elements in the first dimension / axis (rows in the case of 2D array), after the comma in the second dimension (columns for a 2D array).

Here are more interesting cases:

In [67]: a = np.arange(12).reshape(4,3)

In [68]: a
Out[68]:
array([[ 0,  1,  2],
       [ 3,  4,  5],
       [ 6,  7,  8],
       [ 9, 10, 11]])

In [69]: a[:, 1]
Out[69]: array([ 1,  4,  7, 10])

In [70]: a[:, 2]
Out[70]: array([ 2,  5,  8, 11])

In [71]: a[2, :]
Out[71]: array([6, 7, 8])

In [72]: a[[1,2], :]
Out[72]:
array([[3, 4, 5],
       [6, 7, 8]])

Official documentation