machine learning - Why features extraction? -
machine learning - Why features extraction? -
in pattern recognition why of import feature extraction? why have cut down feature space? computational problem or procedure improves generalization ability of classifier?
feature extraction not computational complexity. fixed number of training samples, if number of features becomes sufficiently large, performance of classifier can decrease significantly. see curse of dimensionality.
as practical case, consider classifier using multivariate normal statistics (mean , covariance). n
training samples , k
features, covariance matrix become singular n < k
. therefore, if number of samples cannot increased, necessary cut down number of features in order utilize classifier.
machine-learning feature-extraction
Comments
Post a Comment