4 ms·
When/if used properly, no. The idea behind PCA is to find a set of features with far less dimensionality than the original data. The hope/intent with this sort
by oivey 5y ago
When/if used properly, no. The idea behind PCA is to find a set of features with far less dimensionality than the original data. The hope/intent with this sort of approach is that any more fitted features are just fitting noise.
- hervature 5y agoFor people who are curious, the GP is correct when it comes to fitting the training data. Recall, with enough parameters, we can get 100% on training. The parent’s comment is about testing/validation where we want to avoid overfitting so removing the least important parameters can be helpful.