Local Adaptive and Incremental Gaussian Mixture for Online Density Estimation

摘要

In this paper, we propose an incremental and local adaptive gaussian mixture for online density estimation (LAIM). Using a similarity threshold based criterion, the method is able to allocate components incrementally to accommodate novel data points without affecting previously learned components. A local adaptive learning strategy is presented for estimating density with complex structure in an online way. We also adopt a denoising scheme to make the algorithm more robust to noise. We compared the LAIM to the state-of-art methods for density estimation in both artificial and real data sets, the results show that our method outperforms the compared online counterpart and produces comparable results to the compared batch algorithms.