Array Creation Methods - NumPy Fundamentals

Array Creation Methods

What you'll build

A function create_arrays() that builds five arrays with NumPy's constructors and returns them in a dictionary.

Background

Almost every ML array starts life in one of these constructors: biases start as np.zeros, attention masks as np.ones, np.arange builds index and position vectors, np.linspace builds evenly spaced grids (learning-rate sweeps, plot axes, diffusion noise schedules), and np.eye gives identity matrices and one-hot rows. Knowing which one takes a shape and which takes a range saves a lot of debugging.

Constructors

Call	Result	Notes
np.zeros(shape)	all 0.0	shape is a tuple, e.g. (3, 4); dtype float64
np.ones(shape)	all 1.0	same shape convention
np.arange(stop)	0, 1, ..., stop - 1	half-open: stop is excluded; integer dtype for integer arguments
np.linspace(start, stop, num)	num evenly spaced values	closed: both endpoints included; step = (stop - start) / (num - 1)
np.eye(n)	n x n identity	1.0 on the diagonal, 0.0 elsewhere

Inputs

None. The function takes no arguments.

Expected Return Format

Return a dictionary with these keys:

• 'zeros': array of shape (3, 4) filled with 0.0
• 'ones': array of shape (2, 3) filled with 1.0
• 'arange': the integers 0 to 9, shape (10,)
• 'linspace': 5 evenly spaced values from 0 to 1 inclusive, [0.0, 0.25, 0.5, 0.75, 1.0]
• 'eye': the 4 x 4 identity matrix

Edge cases and conventions

• np.zeros(3, 4) is a bug: the second positional argument is the dtype, so it raises a TypeError. Pass the shape as one tuple.
• np.arange excludes its end point; np.linspace includes it.

Constraints

Use NumPy constructors; don't type the values out as Python lists.

Follow-up

• What does np.linspace(0, 1, 5, endpoint=False) return, and when is that the version you want?
• Why prefer np.linspace over np.arange(0, 1.01, 0.25) when you need a fixed number of float points?

Input: create_arrays()["linspace"]
Output: array([0.  , 0.25, 0.5 , 0.75, 1.  ])

Input: create_arrays()["arange"]
Output: array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])

Input: create_arrays()["zeros"].shape
Output: (3, 4)

import numpy as np

def create_arrays() -> dict:
    """
    Create five arrays with NumPy constructors.

    Returns:
        dict with keys:
            'zeros':    (3, 4) array of 0.0
            'ones':     (2, 3) array of 1.0
            'arange':   integers 0..9, shape (10,)
            'linspace': 5 evenly spaced values from 0 to 1 inclusive
            'eye':      (4, 4) identity matrix
    """
    # Your code here
    pass