SciPy · Class

sparse.csc_matrix

scipy.sparse.csc_matrix creates compressed sparse column matrices. It accepts dense, sparse, coordinate-form, shape-only, and explicit compressed-storage inputs.

How it works

csc_matrix stores a sparse matrix by columns. In its standard representation, each column's row indices occupy a slice of indices selected by adjacent indptr values, while the corresponding numerical entries are in data; it can also be built from dense input, another sparse matrix, dimensions, coordinate triples, or these storage arrays.

Syntax

scipy.sparse.csc_matrix(arg1, shape=None, dtype=None, copy=False)

Parameters

Name	Type	Description
`arg1`	`object`	Constructor input: a 2-D ndarray, sparse array or matrix, shape tuple, coordinate-form tuple, or compressed-form tuple.
`shape`	`tuple[int, int] or None`	Optional matrix shape, written as (M, N).
`dtype`	`dtype or None`	Optional element data type; for an empty matrix it defaults to dtype='d'.
`copy`	`bool`	Whether to copy input data when applicable.

Returns

A scipy.sparse.csc_matrix object.

Key points

csc_matrix represents a matrix in compressed sparse column format.
A two-dimensional NumPy array can be used as the constructor input.
A shape tuple constructs an empty sparse matrix, with an optional dtype.
Coordinate-form construction accepts values together with row and column index arrays.
Each coordinate entry assigns the k-th value to the indexed row and column.
The indices slice bounded by adjacent indptr entries contains the row positions for a column.
The same boundary slice in data contains the values associated with those row positions.

Common mistakes

When constructing from coordinate or compressed arrays, omit shape only when the index information unambiguously determines the intended dimensions.
Passing an existing sparse object converts it to CSC format rather than preserving its original storage format.

Example: Create and inspect a CSC matrix

import numpy as np
from scipy.sparse import csc_matrix

A = csc_matrix(np.array([[1, 0, 2], [0, 3, 0]]))
print(A.toarray())
print(A.indptr)
print(A.indices)
print(A.data)

Creates a CSC matrix from a two-dimensional NumPy array and inspects its dense form and three storage arrays.

When to use it

Memory-efficient sparse storage

Store matrices with many zeros while retaining column-oriented compressed storage.

Coordinate-form assembly

Build a matrix from nonzero values and their row and column coordinates.

Compressed representation control

Construct or inspect explicit CSC arrays for interoperability with sparse algorithms.

Interview questions

QWhat is scipy.sparse.csc_matrix?

It is a compressed sparse column matrix. Its compressed representation uses data for values, indices for row positions, and indptr to delimit each column's segment in the first two arrays.

Follow-up: How do you locate the stored entries of column j? Use indices[indptr[j]:indptr[j+1]] for row positions and data[indptr[j]:indptr[j+1]] for their values.

QWhat input forms does csc_matrix accept?

You can pass a two-dimensional ndarray, another sparse object, a shape tuple, coordinate-form data with row and column indices, or the compressed triple data, indices, indptr. The optional shape specifies matrix dimensions where needed.

Follow-up: What relationship must coordinate-form inputs satisfy? For each k, the entry at row_ind[k], col_ind[k] equals data[k].

QHow does the standard CSC representation encode columns?

For column i, row indices are stored in indices[indptr[i]:indptr[i+1]], and matching values are stored in data[indptr[i]:indptr[i+1]]. Thus indptr provides the boundaries for each column's compressed data.

Follow-up: How can you inspect or validate the representation? Compare each column's boundary slice in indices and data, and ensure indptr has the expected column boundary structure.

scipy.sparse.csr_matrix
scipy.sparse.coo_matrix
scipy.sparse.dia_matrix
scipy.sparse.bsr_matrix

Source: scipy.sparse.csc_matrix. Checked 2026-10-15.