Incremental Slow Feature Analysis
Varun Raj Kompella, Matthew Luciw and Juergen Schmidhuber
The Slow Feature Analysis (SFA) unsupervised learning framework extracts features representing the underlying causes of the changes within a temporally coherent high-dimensional raw sensory input signal. We develop here the first online version of SFA, via a combination of incremental Principal Components Analysis and Minor Components Analysis. The online version allows SFA to adapt along with nonstationary environments, making it useful for autonomous learning. We compare the online SFA to the batch-based SFA in several experiments and show it successfully extracts the slow features given a suitable learning rate. We extend our algorithm to deep networks in hierarchical fashion, and use it to successfully extract abstract object position information from a high-dimensional video stream.