Construct a regular tiled layout
Create repeated rows or columns from an existing array.
NumPy · Function
np.tile
Constructs an array by repeating an input along its axes according to reps. NumPy aligns the input rank and repetition specification by adding leading dimensions or leading repetition counts of one.
How it works
np.tile constructs a new array by repeating the input along each axis according to reps. The number of dimensions is the larger of reps length and the input rank; missing dimensions are handled by prepending axes or repetition counts of one.
Syntax
numpy.tile(A, reps)Parameters
| Name | Type | Description |
|---|---|---|
A | array_like | The input array. |
reps | array_like | The number of repetitions of A along each axis. |
Returns
The tiled output array, as an ndarray.
Key points
- Builds an array containing repeated copies of the input, with repetition controlled by reps.
- The output rank is the larger of the repetition specification length and the input array rank.
- A lower-rank input gains leading singleton axes when reps describes more dimensions.
- For a vector, extra dimensions are added at the front before replication.
- When the input has more axes than reps, leading repetition counts of one are added.
- Each reps entry controls how many times the corresponding axis is repeated.
Common mistakes
- Do not use tile as the default substitute for broadcasting; prefer NumPy broadcasting operations and functions.
Example: Repeat a vector across rows
import numpy as np
values = np.array([1, 2, 3])
result = np.tile(values, (2, 1))
print(result)
# [[1 2 3]
# [1 2 3]]Creates a two-row array by repeating the one-dimensional input once along its existing axis.
When to use it
Replicate multidimensional patterns
Generate repeated blocks for matrix-like data.
Materialize repeated values
Create explicit repeated data when a materialized array is required.
Interview questions
QWhat does np.tile do, and how does it determine the output rank?
np.tile repeats an array according to the counts in reps and returns the tiled output array. If reps has d entries, the result rank is max(d, A.ndim).
Follow-up: How is a one-dimensional input treated with two repetition counts? It is promoted by adding a leading axis, so its shape behaves like (1, n).
QWhat are the parameters of np.tile?
A is the input array-like object, while reps specifies repetitions along each axis. A scalar repetition is accepted, and a sequence supplies one count per represented axis.
Follow-up: What happens when A has more dimensions than reps? Reps is expanded by prepending ones until it matches A.ndim.
QHow does np.tile differ from broadcasting?
The function returns an ndarray containing the repeated input. For example, tiling [0, 1, 2] with (2, 2) produces two rows, each containing the input twice.
Follow-up: When should broadcasting be used instead? Use broadcasting operations when the goal is compatible-shape computation rather than materializing repeated data.
Related
numpy.repeatnumpy.broadcast_to