When does PCA fail?
mediumAnswer
- (1) Non-linear structure (rolls, spirals) — PCA sees only linear correlations.
- (2) Discrete / categorical features (one-hot creates artificial variance).
- (3) Skewed features — one heavy-tail feature dominates; log-transform first.
- (4) Interpretability priority — PC1 mixes many features, hard to name.
- (5) Anomaly-heavy data — outliers distort covariance; use robust PCA.
- Alternatives: KernelPCA, autoencoders, ICA depending on the failure mode.
Check yourself — multiple choice
- Never fails
- Non-linear structure / categoricals / skewed features / interpretability / outliers → use KernelPCA / autoencoders / robust PCA / log-transform
- Same as t-SNE
- Random
PCA fails: non-linear / categorical / skewed / outliers → alternatives.
#dimensionality-reduction#pca
Practise Unsupervised Learning
214 interview questions in this topic.