Streaming Kernel Principal Component Analysis (SKPCA)
In this work, a new explicit kernel map called Explicit Cosine Map was proposed.
By combining this kernel and the βFD algorithm, a fast and accurate SKPCA algorithm was developed.
It was demonstrated to be more accurate than other previous algorithms in practice.
Dataset: PeMS, MNIST, ISOLET, HAR
Implementation done in: Python, Matlab
Software packages used: Scikit-learn, numpy
Preliminary experiments: In this work, various approximate kernel PCA (KPCA) methods such as the Nyström method, randomized nonlinear components analysis,
and random Fourier features (RFF) base streaming KPCA are used as feature extraction methods. The obtained features are then
used for classification tasks. A comparison of classification accuracy is done in order to determine the best approximate KPCA algorithm.