Locality preserving CCA with applications to data visualization and pose estimation

Visualizing Web Images Using Fisher Discriminant Locality Preserving Canonical Correlation Analysis

“…We compared our proposed method (i.e., FDLP-CCA) with seven other dimensionality reduction methods as comparison methods: LPCCA [18], FDA [19], MDS [23], Isomap [8], t-SNE [10], BH-SNE [25] and m-SNE [15]. Since FDLP-CCA is a hybrid version of LPCCA and FDA, LPCCA and FDA correspond to FDLP-CCA of α = 1 and α = 0, respectively.…”

“…By considering the unique characteristics of the text and visual features, FDLP-CCA can not only integrate multiple features maximizing their correlation but also discriminate items as semantics. Specifically, this dimensionality reduction is based on a hybrid version of two dimensionality reduction methods: LPCCA [18] and FDA [19]. This section is organized as follows.…”

“…Specifically, our method is a hybrid version of two dimensionality reduction methods, Locality Preserving Canonical Correlation Analysis (LPCCA) [18] and Fisher Discriminant Analysis (FDA) [19], and we thus call our hybrid method "Fisher Discriminant Locality Preserving Canonical Correlation Analysis" (FDLP-CCA). We formulate the optimization problem of FDLP-CCA by combining those of LPCCA and FDA.…”

Can under-exploited structure of original-classes help ECOC-based multi-class classification?

“…COIL-20 is a database containing gray-scale images of 20 objects, as shown in Fig. 3 [29]. The objects were placed on a motorized turntable against a black background.…”

Low-Resolution Face Recognition via Coupled Locality Preserving Mappings

“…Inspired by the works of [3], [11], we introduce the locality preserving objective into our minimization criterion as follows: (13) Similar with the work of [15], the penalty weighting matrix (with size of ) serves to preserve the local relationship between data points in the original feature spaces.…”