Numpy study notes#

General learning resources#

Show numpy version and the configuration#

np.__version__
np.show_config()

Memory size of an array#

The memory size of an array equals to array.size * array.itemsize

Get the documentation of a numpy function#

  • using np.add as an example:

    np.info(np.add)

    • in command line $ python -c "import numpy as np; np.info33(np.add)"

Compare arange and linspace#

  • Both return evenly spaced numbers over a specified interval.

  • arange is similar to linspace, but it uses step size instead of number of samples

Reverse an vector#

  • np.flip(vec)

  • vec[::-1]

Find non-zero/zero elements in a list#

  • non-zero: np.nonzero(l)

  • zero: np.argwhere(l == 0)

Create identify matrix#

  • np.eye(n)

  • np.identity(n)

  • identity just calls eye so there is no difference in how the arrays are constructed.

  • the main difference is that with eye the diagonal can may be offset, whereas identity only fills the main diagonal.

Pad an array with np.pad#

numpy.pad(array, pad_width, mode='constant', **kwargs)

Understanding np.nan#

numpy.diag#

  • numpy.diag(v, k=0): Extract a diagonal or construct a diagonal array.

    • The default is 0. Use k>0 for diagonals above the main diagonal, and k<0 for diagonals below the main diagonal.

numpy array index#

  • what does the following code mean?

    Z = np.zeros((8,8),dtype=int)
Z[1::2,::2] = 1
Z[::2,1::2] = 1

matrix production in numpy#

  • numpy.dot(m1, m2)

  • m1 @ m2 in Python 3.5 and above

Uncategorized#

  • numpy.unravel_index(99, (6, 7, 8))

  • creating new dtype

    color = np.dtype([("r", np.ubyte),
              ("g", np.ubyte),
              ("b", np.ubyte),
              ("a", np.ubyte)])

Find common values between two array#

  • np.intersect1d(v1, v2), where v1 and v2 are two numpy arrays

Ignore numpy warnings#

  • defaults = np.seterr(all="ignore")

    • back to sanity: np.setarr(**defaults)

  • with a context manager

      with np.errstate(all="ignore"):
      np.arange(3)/0

Date in numpy#

  • today: numpy.datetime64('today')

  • yesterday: numpy.datetime64('today') - numpy.timedelta64(1)

  • get all the dates in cerntain month/between two dates, e.g. np.arange('2016-07', '2016-08', dtype='datetime64[D]')

Doing calculation in place#

  • numpy.add(A, B, out = B)

Check if two arrays are equal#

  • comparing values only: np.allclose(A, B)

  • comapring both shape and values: np.array_equal(A, B)

Make an array immutable#

  • given an array v: v.flags.writeable = False

Read txt file in numpy#

  • Example

    from io import StringIO

# Fake file
s = StringIO('''1, 2, 3, 4, 5

                6,  ,  , 7, 8

                ,  , 9,10,11
            ''')
Z = np.genfromtxt(s, delimiter=",", dtype=np.int)
print(Z)

Sort an array by the nth column#

Z = np.random.randint(0,10,(3,3))
print(Z)
print(Z[Z[:,1].argsort()])

np.reshape()#

  • what does the -1 mean in v.reshape(-1, x)

  • more generally, the index of arraries

np.bincount()#

  • Convert a np.bincount() result back to the original vector

    C = np.bincount([1,1,2,3,4,4,6])
A = np.repeat(np.arange(len(C)), C)
print(A)

np.enisum#

See intro blog here