面向等维独立多流形的增量学习算法IMM-ISOMAP

Manifold learning is an important research direction in the field of machine learning and data mining. The classical algorithm always assumes that the high dimensional batched data exists in a single manifold, and can not effectively deal with the high dimensional manifold data. In this paper, we propose an incremental learning algorithm IMM-ISOMAP for equal dimension independent multi-manifolds. Firstly, the algorithm computes the dynamic neighborhood of the new sample, and divides the new samples into sub-manifolds by extending the tangent space, and then realizes the classification of the new samples and calculates the final low dimensional embedding. Experimental results show that the proposed algorithm can be effectively applied to artificial data and real image data.