Broadcasting with higher dimensions in NumPy - Step-by-Step Execution

Concept Flow - Broadcasting with higher dimensions

Start with arrays of different shapes

↓

Compare shapes from trailing dimensions

↓

Check dimension compatibility

↓

Broadcast smaller array

↓

Perform element-wise operation

↓

Result array

Broadcasting compares array shapes from the end, expands smaller arrays by adding dimensions of size 1, then performs element-wise operations.

Execution Sample

NumPy

import numpy as np
A = np.array([[1, 2, 3], [4, 5, 6]])
B = np.array([10, 20, 30])
C = A + B[:, np.newaxis]

Add a 1D array B to a 2D array A by reshaping B to match A's dimensions using broadcasting.

Execution Table

Step	A shape	B shape	B reshaped shape	Broadcasting check	Operation	Result shape
1	(2, 3)	(3,)	(3, 1)	Compare trailing dims: 3 vs 1 (compatible)	Broadcast B to (3,1)	(2, 3) vs (3, 1) shapes not aligned yet
2	(2, 3)	(3, 1)	Broadcast B to (2, 3)	Expand B dims to match A	Add A + B broadcasted	(2, 3)
3	(2, 3)	(3,)	(3, 1)	Final broadcasted shapes: A (2,3), B (2,3)	Element-wise addition	(2, 3)
4	(2, 3)	(3,)	(3, 1)	Result array created	Output C	(2, 3)

💡 Broadcasting completed successfully; result shape is (2, 3)

Variable Tracker

Variable	Start	After Step 1	After Step 2	Final
A	array([[1, 2, 3], [4, 5, 6]])	Same	Same	Same
B	array([10, 20, 30])	Reshaped to (3,1)	Broadcasted to (2,3)	Broadcasted to (2,3)
C	Not defined	Not defined	Computed as A + B broadcasted	array([[11, 12, 13], [24, 25, 26]])

Key Moments - 3 Insights

Why do we reshape B to (3,1) before adding to A?

Reshaping B to (3,1) adds a new axis so its dimensions align properly with A's (2,3) shape for broadcasting, as shown in execution_table step 1.

How does broadcasting handle different shapes like (2,3) and (3,1)?

Broadcasting compares shapes from the right; dimensions of size 1 can be stretched to match the other array, enabling element-wise operations as in execution_table steps 2 and 3.

What happens if shapes are not compatible for broadcasting?

If dimensions differ and neither is 1, broadcasting fails with an error. This is shown in the concept_flow where incompatible shapes lead to an error.

Visual Quiz - 3 Questions

Look at the execution_table at step 1, what is the shape of B after reshaping?

A (3, 1)

B (1, 3)

C (2, 3)

D (3,)

💡 Hint

Check the 'B reshaped shape' column in execution_table row for step 1.

At which step does the final broadcasted shape of B match A's shape?

A Step 1

B Step 2

C Step 3

D Step 4

💡 Hint

Look at the 'Broadcasting check' and 'Result shape' columns in execution_table.

If B was not reshaped before addition, what would happen?

A Broadcasting would fail with an error

B B would automatically reshape to (2,3)

C The result would have shape (3,)

D The operation would produce a scalar

💡 Hint

Refer to concept_flow where incompatible shapes cause an error.

Concept Snapshot

Broadcasting with higher dimensions:
- Compare shapes from trailing dimensions
- Dimensions must be equal or one must be 1
- Smaller arrays are reshaped by adding axes
- Arrays expand to compatible shapes
- Element-wise operations then apply
- Result shape is the broadcasted shape

Full Transcript

Broadcasting with higher dimensions means adding arrays of different shapes by expanding the smaller array's dimensions. We compare shapes from the right side. If dimensions match or one is 1, we can stretch the smaller array. For example, adding a (2,3) array to a (3,) array requires reshaping the smaller array to (3,1) first. Then broadcasting expands it to (2,3). The addition happens element-wise, producing a (2,3) result. If shapes are not compatible, broadcasting fails with an error.