DGPAGE | R Documentation |
This package-like script provides a complete, faithful, working out-of-the-box R implementation of the DGPAGE method described in:
J. Gou, X. Yuan, Y. Xue, L. Du, J. Yu, S. Xia, Y. Zhang."Discriminative and Geometry-Preserving Adaptive Graph Embedding for dimensionality reduction."Information Sciences, 2022.
DGPAGE systematically integrates:
Manually constructed similarity and diversity graphs (see Eqs. (4) and (7) in the paper),
An adaptively learned adjacency graph with diversity-based regularization (Eqs. (8)-(11)),
A multi-task-like objective that jointly optimizes the adjacency matrix and the projection matrix (Eq. (12)), and
An alternating optimization algorithm (Algorithm 1 in the paper) that solves for these parameters.
The main user-facing functions are:
dgpage_fit()
: fits the DGPAGE model on training data,
dgpage_predict()
: obtains predictions (via 1-NN) on new test data,
compute_similarity_graph()
: constructs the manual similarity graph G_S
(Eq. (4)),
compute_diversity_graph()
: constructs the manual diversity graph G_D
(Eq. (7)).
The code below relies on base R (for matrix operations) and uses the Rnanoflann package for 1-nearest-neighbor classification.
Note: For large datasets n
, the O(n^2)
graph construction may
be expensive in plain R. You might want to accelerate via Rcpp or other means.
Original Paper: Jianping Gou, Xia Yuan, Ya Xue, Lan Du, Jiali Yu, Shuyin Xia, Yi Zhang
R Implementation: Your Name (youremail@domain.com)
Gou, J., Yuan, X., Xue, Y., Du, L., Yu, J., Xia, S., & Zhang, Y. (2022). Discriminative and Geometry-Preserving Adaptive Graph Embedding for dimensionality reduction. Information Sciences.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.