Why sparse matrices save memory in SciPy - Challenge Your Understanding

Challenge - 5 Problems

🧠 Conceptual

intermediate

Why do sparse matrices save memory?

Imagine you have a large matrix mostly filled with zeros. Why does using a sparse matrix save memory compared to a regular dense matrix?

A Sparse matrices compress all elements including zeros to save space.

B Sparse matrices store only the non-zero elements and their positions, reducing memory use.

C Sparse matrices convert zeros to ones to reduce memory.

D Sparse matrices store the entire matrix twice for faster access.

💡 Hint

Think about what data is actually stored in a sparse matrix.

Sparse matrices save memory by storing only the non-zero values and their locations, instead of storing every element including zeros.

❓ Predict Output

intermediate

Memory usage of dense vs sparse matrix

What is the output of this code showing memory usage in bytes?

SciPy

import numpy as np
from scipy.sparse import csr_matrix

size = 10000
matrix_dense = np.zeros((size, size))
matrix_dense[0, 0] = 1
matrix_sparse = csr_matrix(matrix_dense)

print(matrix_dense.nbytes)
print(matrix_sparse.data.nbytes + matrix_sparse.indptr.nbytes + matrix_sparse.indices.nbytes)

A

800000000
80000

B

800000000
800000000

C

80000
800000000

D

80000
80000

💡 Hint

Check how many bytes the dense matrix uses and compare to the sparse matrix components.

The dense matrix stores all elements (10000x10000 floats), using 800 million bytes. The sparse matrix stores only non-zero data and index arrays, much smaller.

❓ data_output

advanced

Number of stored elements in sparse matrix

Given a 5x5 matrix with only 3 non-zero elements, how many elements does the sparse matrix store?

SciPy

from scipy.sparse import csr_matrix
import numpy as np

matrix = np.zeros((5,5))
matrix[0,1] = 10
matrix[2,3] = 20
matrix[4,4] = 30
sparse = csr_matrix(matrix)

print(len(sparse.data))

A 5

B 25

C 0

D 3

💡 Hint

Count how many non-zero values are in the matrix.

The sparse matrix stores only the non-zero values. Here, there are exactly 3 non-zero elements.

🔧 Debug

advanced

Why does this sparse matrix code raise an error?

What error does this code raise and why?

SciPy

from scipy.sparse import csr_matrix

matrix = [[0, 0], [0, 0]]
sparse = csr_matrix(matrix)
print(sparse.data[0])

A IndexError because sparse.data is empty (no non-zero elements).

B TypeError because matrix is a list, not numpy array.

C ValueError because matrix shape is invalid.

D AttributeError because csr_matrix has no attribute 'data'.

💡 Hint

Check what happens when the matrix has no non-zero elements.

The sparse matrix has no non-zero elements, so sparse.data is empty. Accessing sparse.data[0] causes IndexError.

🚀 Application

expert

Choosing sparse matrix format for memory efficiency

You have a large sparse matrix with many rows but few non-zero elements per row. Which sparse format is best to save memory and why?

A DOK (Dictionary of Keys) because it uses a dictionary and is memory efficient for large matrices.

B CSC (Compressed Sparse Column) because it stores column indices and is better for column slicing.

C CSR (Compressed Sparse Row) because it efficiently stores row indices and data for fast row slicing.

D LIL (List of Lists) because it stores data as lists and is best for memory.

💡 Hint

Think about which format stores data efficiently for many rows with few non-zero elements each.

CSR format stores data efficiently by compressing rows and storing only non-zero elements and their column indices, saving memory for matrices with many rows and few non-zero elements per row.