Matrix factorization basics in ML Python - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is matrix factorization in simple terms?

Matrix factorization is a way to break a big table of numbers into two smaller tables that, when multiplied, give back the original table. It helps find hidden patterns.

beginner

Why do we use matrix factorization in machine learning?

We use matrix factorization to discover hidden features or patterns in data, reduce data size, and make predictions, like recommending movies based on user ratings.

beginner

What are the two smaller matrices called in matrix factorization?

They are often called the 'user matrix' and the 'item matrix' in recommendation systems, or more generally, two factor matrices that multiply to approximate the original matrix.

intermediate

How does matrix factorization help in recommendation systems?

It finds hidden features of users and items by breaking down the rating matrix, so it can predict how much a user might like an item they haven't seen yet.

intermediate

What is the role of the 'rank' in matrix factorization?

The rank is the number of hidden features or factors we choose. It controls how detailed the approximation is: too low loses info, too high may overfit.

What does matrix factorization do to a large matrix?

A Changes it into text

B Converts it into a single number

C Deletes half of the data

D Breaks it into two smaller matrices

Matrix factorization breaks a large matrix into two smaller matrices whose product approximates the original.

In recommendation systems, what do the two factor matrices represent?

A Days and months

B Colors and shapes

C Users and items

D Prices and discounts

The two matrices represent users and items, capturing their hidden features.

What is a common goal of matrix factorization in machine learning?

A Predict missing values

B Increase matrix size

C Remove all zeros

D Sort data alphabetically

Matrix factorization helps predict missing values by learning hidden patterns.

What happens if the rank chosen for factorization is too low?

A The model runs faster but is perfect

B Important details may be lost

C The matrix becomes larger

D The matrix turns into zeros

A low rank means fewer features, which can miss important information.

Which of these is NOT a benefit of matrix factorization?

A Encrypting data securely

B Finding hidden patterns

C Making predictions

D Reducing data size

Matrix factorization does not encrypt data; it helps find patterns and reduce size.

Explain matrix factorization and why it is useful in machine learning.

Think about how big data tables can be simplified to find patterns.

Describe the role of the rank in matrix factorization and its impact on the model.

Consider how choosing too few or too many features affects results.