How to Create a Sparse Matrix with SciPy

A sparse matrix is a matrix in which most elements are zeros. Sparse matrices are widely used in machine learning, natural language processing (NLP), and large-scale data processing, where storing all zero values is inefficient.

Example of a sparse matrix:

0 0 3 0 4
0 0 5 7 0
0 0 0 0 0
0 2 6 0 0

Storing such a matrix as a normal 2D array wastes memory, as most elements are zeros. Instead, we store only non-zero elements along with their row and column indices (triplets format).

Benefits of using sparse matrices:

• Reduced Memory Usage: Only non-zero elements are stored, saving memory.
• Faster Computations: Operations can be performed only on non-zero elements, improving speed.

Sparse Matrix Formats in SciPy

The scipy.sparse module provides several formats for storing sparse matrices, each optimized for different operations:

Example 1: csr_matrix (Compressed Sparse Row)

CSR format stores non-zero values row-wise, enabling fast row slicing and efficient matrix operations.

Output

[[0 0 3 0 4]
[0 0 5 7 0]
[0 0 0 0 0]
[0 2 6 0 0]]

Explanation: csr_matrix stores only non-zero values with their coordinates and reconstructs full matrix using toarray().

Example 2: csc_matrix (Compressed Sparse Column)

CSC format stores data column-wise, making column-based operations faster.

Output

[[0 0 3 0 4]
[0 0 5 7 0]
[0 0 0 0 0]
[0 2 6 0 0]]

Explanation: Stores non-zero values in column-compressed format, efficient for column operations.

Example 3: coo_matrix (Coordinate Format)

COO format represents the matrix using (row, col, value) triplets. Useful when constructing matrices dynamically before converting to CSR/CSC.

Output

[[0 0 3 0 4]
[0 0 5 7 0]
[0 0 0 0 0]
[0 2 6 0 0]]

Explanation: Stores elements as (row, col, value) tuples.

Example 4: lil_matrix (List of Lists)

LIL (List of Lists) format allows efficient row-wise construction. You can easily insert or modify values before converting the matrix to CSR or CSC for faster computation.

Output

[[0. 0. 3. 0. 4.]
[0. 0. 5. 7. 0.]
[0. 0. 0. 0. 0.]
[0. 2. 6. 0. 0.]]

Explanation: Creates a List of Lists (LIL) matrix and assigns values directly by row and column.

Example 5: dok_matrix (Dictionary of Keys)

DOK (Dictionary of Keys) format is ideal for random assignments. You can assign elements at any position efficiently, making it perfect for incremental matrix construction.

Output

[[0. 0. 3. 0. 4.]
[0. 0. 5. 7. 0.]
[0. 0. 0. 0. 0.]
[0. 2. 6. 0. 0.]]

Explanation: Internally stored as dictionary {(row, col): value}.

Example 6: dia_matrix (Diagonal Matrix)

DIA (Diagonal) format stores only the diagonals of the matrix. It is very memory-efficient for diagonal-dominant matrices, where most non-zero elements lie along certain diagonals.

Output

[[3 0 0 0 0]
[0 5 0 0 0]
[0 0 6 0 0]
[0 0 0 7 0]]

Explanation: Creates a Diagonal (DIA) matrix storing only specified diagonals.

Related Articles:

• Compressed Sparse formats CSR and CSC in Python
• Python program to Convert a Matrix to Sparse Matrix