Learning Association Relationship and Accurate Geometric Structures for Multi-Type Relational Data

Learning Inter- and Intra-manifolds for Matrix Factorization-based Multi-Aspect Data Clustering

“…The consensus manifold is further exploited by the novel optimal manifold concept proposed in 2CMV [30] which embeds the most consensed manifold in multi-view data. Recently, ARASP [11] proposed a novel fashion of learning the MTRD manifold, where the close and far distance information is embedded and preserved steadily for each data type. These works have proved the importance and necessity of manifold learning with clustering.…”

“…Existing methods using the manifold learning in MTRD and multi-view data focus on learning the intra-manifold within each data type or within data samples in each data view respectively [11][17] [18]. These methods preserve local geometric structure or close distances between the data points of each data type or each view only.…”

“…The real-world data is known to lie on multiple man- ifolds embedded in the high and multi dimensional data space [9][10] [11]. Manifold learning has been applied to preserve the geometric structure of original data when projecting from high to lower dimensional space [12].…”

“…Manifold learning has been incorporated in NMF to learn the partbased representation that respects the intrinsic manifold of data and assists in obtaining more accurate clusters [2][13][14] [15]. Many efforts have been made to learn the accurate manifold of MTRD [3] [11][16] [17]. These methods focus on learning the manifold on each object type explicitly, i.e., considering how data points of the same object type reside on the manifold.…”