Truly Incremental Locally Linear Embedding
Sebastian Schuon, Marko Đurković, Klaus J. Diepold, Jürgen Scheuerle, Stefan Markward · 2008
Abstract—We describe an incremental version of the Locally Linear Embedding (LLE) method for manifold learning. Such an incremental LLE algorithm is suitable for real time application, where measurement data is coming in continuously. We are able to incrementally calculate embeddings which provide the same level of accuracy as the original LLE algorithm. For situations where data is appended constantly, the proposed algorithm achieves a significant speed up. I.