Sparse SVD (svds) in SciPy - Mini Project: Build & Apply

Sparse SVD with svds from SciPy

📖 Scenario: You work as a data analyst for a movie streaming service. You have a large matrix showing user ratings for movies, but most users have rated only a few movies, so the matrix is mostly empty (sparse). You want to find patterns in this data using a technique called Sparse Singular Value Decomposition (Sparse SVD).

🎯 Goal: Build a Python program that creates a sparse matrix of user ratings, configures the number of singular values to find, applies Sparse SVD using svds from SciPy, and prints the singular values.

📋 What You'll Learn

Create a sparse matrix using scipy.sparse.csr_matrix with given data

Set a variable k for the number of singular values to compute

Use svds from scipy.sparse.linalg to compute the sparse SVD

Print the singular values array

💡 Why This Matters

🌍 Real World

Sparse SVD is used in recommendation systems to find hidden patterns in large, sparse user-item rating data.

💼 Career

Data scientists and machine learning engineers use sparse matrix decompositions to reduce data size and improve model performance.

1

Create a sparse matrix of user ratings

Create a sparse matrix called ratings using scipy.sparse.csr_matrix with the exact data: rows = [0, 0, 1, 2, 2], cols = [0, 2, 2, 0, 1], and data = [5, 3, 4, 1, 2].

SciPy

from scipy.sparse import csr_matrix
# Create the sparse matrix 'ratings' with given rows, cols, and data
# Your code here

Need a hint?

Use csr_matrix((data, (rows, cols)), shape=(3, 3)) to create the sparse matrix.

2

Set the number of singular values to compute

Create a variable called k and set it to 2 to specify the number of singular values to compute.

SciPy

from scipy.sparse import csr_matrix
rows = [0, 0, 1, 2, 2]
cols = [0, 2, 2, 0, 1]
data = [5, 3, 4, 1, 2]
ratings = csr_matrix((data, (rows, cols)), shape=(3, 3))
# Set the variable 'k' to 2
# Your code here

Need a hint?

Just write k = 2 to set the number of singular values.

3

Compute Sparse SVD using svds

Import svds from scipy.sparse.linalg and use it to compute the sparse SVD of ratings with k singular values. Store the results in variables u, s, and vt.

SciPy

from scipy.sparse import csr_matrix
rows = [0, 0, 1, 2, 2]
cols = [0, 2, 2, 0, 1]
data = [5, 3, 4, 1, 2]
ratings = csr_matrix((data, (rows, cols)), shape=(3, 3))
k = 2
# Import svds and compute u, s, vt from ratings with k
# Your code here

Need a hint?

Use from scipy.sparse.linalg import svds and then u, s, vt = svds(ratings, k=k).

4

Print the singular values

Write a print statement to display the singular values stored in the variable s.

SciPy

from scipy.sparse import csr_matrix
from scipy.sparse.linalg import svds
rows = [0, 0, 1, 2, 2]
cols = [0, 2, 2, 0, 1]
data = [5, 3, 4, 1, 2]
ratings = csr_matrix((data, (rows, cols)), shape=(3, 3))
k = 2
u, s, vt = svds(ratings, k=k)
# Print the singular values 's'
# Your code here

Need a hint?

Use print(s) to show the singular values.