NumPy Arrays

The ndarray is NumPy's core data structure — a grid of values, all of the same type, with a fixed shape. Understanding shape and indexing is the single most useful NumPy skill for ML.

Creating Arrays

import numpy as np

a = np.array([1, 2, 3])                  # 1D array, shape (3,)
b = np.array([[1, 2], [3, 4], [5, 6]])   # 2D array, shape (3, 2) — 3 rows, 2 columns

zeros = np.zeros((2, 3))                 # 2x3 array of zeros
ones  = np.ones((3, 3))                  # 3x3 array of ones
rng   = np.arange(0, 10, 2)              # [0 2 4 6 8]
lin   = np.linspace(0, 1, 5)             # [0. 0.25 0.5 0.75 1.] — 5 evenly spaced points

Shape and dtype

X = np.array([[1200, 3], [800, 2], [1500, 4]])
print(X.shape)   # (3, 2) — 3 samples, 2 features. This is the shape sklearn expects for X.
print(X.dtype)   # int64 — every element is the same type; unlike Python lists

In ML code, X.shape is checked constantly — a shape mismatch between your features and what a model expects is one of the most common runtime errors, and reading the error's expected-vs-actual shape is usually enough to debug it.

Indexing and Slicing

X = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]])

print(X[0])        # [1 2 3] — first row
print(X[:, 0])      # [1 4 7] — first column, ALL rows
print(X[0:2, 1:3])  # [[2 3], [5 6]] — sub-matrix
print(X[X > 5])     # [6 7 8 9] — boolean/fancy indexing: all values > 5, flattened

Reshaping

flat = np.array([1, 2, 3, 4, 5, 6])
matrix = flat.reshape(2, 3)     # reorganize into 2 rows, 3 columns
print(matrix)
# [[1 2 3]
#  [4 5 6]]

col = np.array([1, 2, 3]).reshape(-1, 1)   # common pattern: turn a 1D array into a column
print(col.shape)   # (3, 1) — scikit-learn often requires this shape for a single feature

The -1 tells NumPy "figure out this dimension automatically" — reshape(-1, 1) is extremely common in ML code when scikit-learn expects a 2D array but you have a 1D one.

Practical Use Cases

• Reshaping a single feature column into the 2D shape scikit-learn requires
• Slicing out training batches, features, or specific rows/columns from a dataset
• Boolean masking to filter samples matching a condition (e.g. outlier removal)

Common Mistakes

• Passing a 1D array where scikit-learn expects a 2D array (a ValueError: Expected 2D array, got 1D array instead) — fix with .reshape(-1, 1).
• Confusing X[0] (first row of a 2D array) with X[:, 0] (first column, all rows) — one of the most common indexing slips.

Interview Relevance

Q: "You get 'Expected 2D array, got 1D array instead' from scikit-learn — what's happening and how do you fix it?" scikit-learn models expect X with shape (n_samples, n_features), even for a single feature — reshape a 1D array with .reshape(-1, 1) to add the required second dimension.

Practice Question

Given a 1D NumPy array of 12 values, write the code to reshape it into a 2D array with 3 rows and 4 columns, and separately into a column vector of shape (12, 1).

Related ML Notes

• NumPy for ML
• Pandas DataFrame
• scikit-learn