The result of PCA is a linear projection matrix. Applying that projection is fast. You learn the projection from an initial dataset. That is "slow". You appl

The result of PCA is a linear projection matrix. Applying that projection is fast. You learn the projection from an initial dataset. That is "slow". You apply the projection to new data. That is "fast".