Two-dimensional supervised local similarity and diversity projection

Globalized and localized matrix-pattern-oriented classification machine

“…Secondly, the matrix-oriented method using structural information in feature extraction has been widely developed in recent years. To improve the traditional Two-Dimensional LPP (2DLPP), the Two-Dimensional Supervised Local Similarity and Diversity Projection (2DSLSDP) is proposed through defining two novel weighted adjacency graphs named similarity graph and diversity graph, respectively [9]. An extension of graph-based image feature extraction method named Sparse Two-Dimensional Locality Discriminant Projections (S2DLDP) is proposed to reduce the high computational cost of the traditional 2DLGEDA and 2DDLPP [15].…”

Joint geometry and variability for image recognition

“…Many previous works have demonstrated that both local geometrical structure and local variation of data are very important for dimensionality reduction and image classification [5,[8][9][10][11][12]. The local geometrical structure of a training data set may be captured by an adjacency graph of data.…”

“…Therefore, they cannot guarantee that the larger the distance between nearby two points is, the farther they should be embedded in the reduced space, creating distortions in the local geometry. Moreover, the local geometrical structure preserved by them characterizes only the locality of data [8] and ignores the variation of the local neighborhoods in the reduced space [9,12].…”

“…Weinberger and Saul pointed out that the local variation of the values among nearby images characterizes the most important modes of variability of data [9,10] and is very important for data representation and classification [9,10,12,29,30]. Moreover, by maximizing the variance, i.e.…”

(2D)2LDALPP: A Novel Approach to Face Recognition

“…��D� � LDA is a supervised learning method, but some research indicates that it is ambiguous on computing because there are different ways to construct the covariance matrix in the��D� � LDA approach. The second type of method pays attention to the algorithm itself, which tries to amend 2D techniques by combining other optimal methods such as [14] and [15].These methods improve the performance of recognition by adding optimal methods to 2D techniques; at the same time, the design of algorithm complexity is also improved. Recently, some methods that hybridize different feature extraction techniques have become popular.…”