Interview: 'when should you use PCA vs autoencoder for dim reduction?'
mediumAnswer
- PCA: (1) linear structure, (2) < 200 features, (3) want interpretability (loadings), (4) fast + closed form, (5) baseline.
- Autoencoder: (1) non-linear structure (images, sequences), (2) very high dim (thousands+), (3) enough data to train, (4) willing to accept less interpretability, (5) transformable to new points, (6) regularizable (denoising, sparsity).
- Rule: start with PCA — if downstream metric plateaus, try autoencoder to catch non-linear info.
Check yourself — multiple choice
- Random
- PCA: linear / small features / interpretable / fast baseline; AE: non-linear / high-dim / lots of data / less interpretable but flexible; start with PCA, escalate if needed
- Only PCA
- Same thing
PCA vs AE: PCA linear/small/interpretable baseline; AE non-linear/high-dim/flexible.
#interview#dimensionality-reduction
Practise Unsupervised Learning
214 interview questions in this topic.
Related questions
- Interview: 'you have 500 features, most correlated — how do you preprocess?'
- PCA vs t-SNE vs UMAP — when do you use each?
- How do you decide how many PCA components to keep?
- Why does clustering degrade in high dimensions?
- Interview: 'you're asked to segment 10M customers — how?'
- Derive PCA — what does it optimize?