Deep supervised t-SNE for SAR target recognition

Yu, Meiting; Zhang, Siqian; Zhao, Lingjun; Kuang, Gangyao

doi:10.1109/icfst.2017.8210516

Cited by 10 publications

(9 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, Yu et al [12] and Cheng et al [13] proposed supervised versions of t-SNE. In this method, the distances between the different classes are calculated as follows.…”

Section: S-isomap=mentioning

confidence: 99%

“…where θ( ) refers to the angle information [12] and the silhouette frame information [13] of the sample .…”

Section: St-mentioning

confidence: 99%

“…This fact makes the use of dissimilarity measure more appropriate in classification problems than the Euclidean distance. The advantage of the proposed S-tSNE approach over the Yu et al [12] and Cheng et al [13] approaches is that it can be applied to any high dimensional dataset. Consequently, the S-tSNE approach can be considered as a generalized case of the other supervised t-SNE approaches.…”

Section: St-mentioning

confidence: 99%

“…Original Data t-SNE S-tSNE MNIST(5,000) 95.30 % 74.00 % 100 % Chest Xray (5,614) 59.10 % 57.27 % 96.32 % SEER Breast Cancer (12,000) 81.86 % 77.25 % 98.81%…”

Section: Datasetmentioning

confidence: 99%

“…Supervised approaches such as WeightedIso, S-Isomap, Supervised LLE, Supervised Enhanced LLE [8,9,10,11] improved the classification accuracy by introducing different kinds of dissimilarity measures instead of using the Euclidean distance. Similarly, Yu et al [12] and Cheng et al [13] proposed supervised t-SNE methods using the class information combined with specific dataset related information (angle information of the Synthetic Aperture Radar (SAR) dataset) [12] and (silhouette frame information of IXMAS and Weizmann datasets) [13]. Consequently, these methods are closely tied to datasets and cannot be generalized to other datasets which do not include the angular or silhouette frame information.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

A New Supervised t-SNE with Dissimilarity Measure for Effective Data Visualization and Classification

Hajderanj

Weheliye

Chen

2019

Proceedings of the 2019 8th International Conference on Software and Information Engineering

View full text Add to dashboard Cite

In this paper, a new version of the Supervised t-Stochastic Neighbor Embedding (S-tSNE) algorithm is proposed which introduces the use of a dissimilarity measure related to class information. The proposed S-tSNE can be applied in any high dimensional dataset for visualization or as a feature extraction for classification problems. In this study, the S-tSNE is applied to three datasets MNIST, Chest x-ray, and SEER Breast Cancer. The two-dimensional data generated by the S-tSNE showed better visualization and an improvement in terms of classification accuracy in comparison to the original t-Stochastic Neighbor Embedding(t-SNE) method. The results from k-nearest neighbors (k-NN) classification model which used the lower dimension space generated by the new S-tSNE method showed more than 20% improvement on average in accuracy in all the three datasets compared with the t-SNE method. In addition, the classification accuracy using the S-tSNE for feature extraction was even higher than classification accuracy obtained from the original high dimensional data.

show abstract

“…In addition, Yu et al [12] and Cheng et al [13] proposed supervised versions of t-SNE. In this method, the distances between the different classes are calculated as follows.…”

Section: S-isomap=mentioning

confidence: 99%

“…where θ( ) refers to the angle information [12] and the silhouette frame information [13] of the sample .…”

Section: St-mentioning

confidence: 99%

Section: St-mentioning

confidence: 99%

“…Original Data t-SNE S-tSNE MNIST(5,000) 95.30 % 74.00 % 100 % Chest Xray (5,614) 59.10 % 57.27 % 96.32 % SEER Breast Cancer (12,000) 81.86 % 77.25 % 98.81%…”

Section: Datasetmentioning

confidence: 99%