Singular Value Decomposition (SVD) — Concept Check
Questions on Singular Value Decomposition (SVD). SVD factors ANY matrix into rotate → scale → rotate (U Σ Vᵀ), giving you the single most useful tool for compression, noise removal, recommendations, and low-rank model tricks like LoRA.
Questions in this quiz
8 questions on Singular Value Decomposition (SVD) — Concept Check. Work through them here, then start the quiz to answer them and see your score.
1.A data scientist is using SVD for image compression. After decomposing the image matrix, they decide to keep only the top 10 singular values. What is the effect of this decision?
- AThe image quality is reduced but storage requirements decrease.
- BThe image quality improves while storage requirements decrease.
- CThe image quality improves with increased storage requirements.
- DThe image quality and storage requirements remain unchanged.
2.A machine learning engineer is using SVD to fine-tune a large language model using LoRA. What role does SVD play in this process?
- AIt identifies important directions for efficient parameter updates.
- BIt increases the rank of the model for better performance.
- CIt eliminates the need for backpropagation in training.
- DIt ensures the model weights are initialized randomly.
3.When performing SVD on a 6x4 matrix, what are the dimensions of the resulting U, Σ, and Vᵀ matrices?
- AU: 6x4, Σ: 4x4, Vᵀ: 4x4
- BU: 6x4, Σ: 6x6, Vᵀ: 4x6
- CU: 6x4, Σ: 4x4, Vᵀ: 6x4
- DU: 6x6, Σ: 6x4, Vᵀ: 4x4
4.A developer is tasked with reducing noise in a dataset using SVD. Which step should they take after performing SVD?
- AZero out the smallest singular values.
- BScale the matrix using U.
- CRotate the matrix using Vᵀ.
- DIncrease the number of singular values.
5.An engineer notices that their recommendation system is slow. They decide to use SVD for optimization. What is the primary benefit of applying SVD here?
- AIt increases the rank of the matrix, improving accuracy.
- BIt ensures all matrix operations are parallelized.
- CIt reduces the size of the user-item matrix to zero.
- DIt provides a low-rank approximation, speeding up computations.
6.A student is reconstructing a matrix using SVD. They mistakenly believe that U and Vᵀ are interchangeable. Why is this incorrect?
- AU and Vᵀ are both scaling matrices.
- BU and Vᵀ are both identity matrices.
- CU and Vᵀ have different roles in rotation and cannot be swapped.
- DU and Vᵀ are both diagonal matrices.
7.What does the singular value in SVD represent in the context of matrix decomposition?
- AThe number of columns in the matrix.
- BThe size of the original matrix.
- CThe amount of 'energy' or importance of each direction.
- DThe number of rows in the matrix.
8.A data analyst uses SVD to approximate a matrix with rank 3. They keep only the top 3 singular values. What is the likely outcome?
- AThe matrix becomes a zero matrix.
- BThe matrix is approximated with minimal loss of information.
- CThe matrix is perfectly reconstructed without approximation.
- DThe matrix dimensions are reduced to 3x3.