Semi-Supervised L2KC (S-L2KC) Classifier

Yue Lv
Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing, PMLR 278:519-526, 2025.

Abstract

Building upon the density difference paradigm, a novel kernel classifier distinct from Support Vector Machines (SVM) — the L2-norm Kernel Classifier (L2KC) — has been developed. This methodology establishes an integrated squared error(ISE) criterion to estimate the true ${{d}_{\gamma }}\left( x \right) $ through minimizing the L2-distance between ${{d}_{\gamma }}\left( x \right) $ and ${{\overset{\scriptscriptstyle\frown}{d}}_{\gamma }}\left( x \right) $, thereby achieving classification via explicit density difference representation. While L2KC demonstrates comparable accuracy to SVM with enhanced decision efficiency, its performance on real-world semi-supervised datasets requires improvement. To address this limitation, we propose the Semi-supervised L2KC (S-L2KC) by incorporating a locality-preserving projection (LPP) based manifold regularization term into the L2KC objective function. This integration effectively enforces the manifold assumption. Experimental results on benchmark datasets from the UCI and LIBSVM demonstrate that compared to L2KC, the proposed S-L2KC exhibits superior generalization capability, characterized by higher mean test accuracy with comparable or even smaller variance.

Cite this Paper

BibTeX
@InProceedings{pmlr-v278-lv25a, title = {Semi-Supervised L2KC (S-L2KC) Classifier}, author = {Lv, Yue}, booktitle = {Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing}, pages = {519--526}, year = {2025}, editor = {Zeng, Nianyin and Pachori, Ram Bilas and Wang, Dongshu}, volume = {278}, series = {Proceedings of Machine Learning Research}, month = {25--27 Apr}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v278/main/assets/lv25a/lv25a.pdf}, url = {https://proceedings.mlr.press/v278/lv25a.html}, abstract = {Building upon the density difference paradigm, a novel kernel classifier distinct from Support Vector Machines (SVM) — the L2-norm Kernel Classifier (L2KC) — has been developed. This methodology establishes an integrated squared error(ISE) criterion to estimate the true ${{d}_{\gamma }}\left( x \right) $ through minimizing the L2-distance between ${{d}_{\gamma }}\left( x \right) $ and ${{\overset{\scriptscriptstyle\frown}{d}}_{\gamma }}\left( x \right) $, thereby achieving classification via explicit density difference representation. While L2KC demonstrates comparable accuracy to SVM with enhanced decision efficiency, its performance on real-world semi-supervised datasets requires improvement. To address this limitation, we propose the Semi-supervised L2KC (S-L2KC) by incorporating a locality-preserving projection (LPP) based manifold regularization term into the L2KC objective function. This integration effectively enforces the manifold assumption. Experimental results on benchmark datasets from the UCI and LIBSVM demonstrate that compared to L2KC, the proposed S-L2KC exhibits superior generalization capability, characterized by higher mean test accuracy with comparable or even smaller variance.} }
Endnote
%0 Conference Paper %T Semi-Supervised L2KC (S-L2KC) Classifier %A Yue Lv %B Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing %C Proceedings of Machine Learning Research %D 2025 %E Nianyin Zeng %E Ram Bilas Pachori %E Dongshu Wang %F pmlr-v278-lv25a %I PMLR %P 519--526 %U https://proceedings.mlr.press/v278/lv25a.html %V 278 %X Building upon the density difference paradigm, a novel kernel classifier distinct from Support Vector Machines (SVM) — the L2-norm Kernel Classifier (L2KC) — has been developed. This methodology establishes an integrated squared error(ISE) criterion to estimate the true ${{d}_{\gamma }}\left( x \right) $ through minimizing the L2-distance between ${{d}_{\gamma }}\left( x \right) $ and ${{\overset{\scriptscriptstyle\frown}{d}}_{\gamma }}\left( x \right) $, thereby achieving classification via explicit density difference representation. While L2KC demonstrates comparable accuracy to SVM with enhanced decision efficiency, its performance on real-world semi-supervised datasets requires improvement. To address this limitation, we propose the Semi-supervised L2KC (S-L2KC) by incorporating a locality-preserving projection (LPP) based manifold regularization term into the L2KC objective function. This integration effectively enforces the manifold assumption. Experimental results on benchmark datasets from the UCI and LIBSVM demonstrate that compared to L2KC, the proposed S-L2KC exhibits superior generalization capability, characterized by higher mean test accuracy with comparable or even smaller variance.
APA
Lv, Y.. (2025). Semi-Supervised L2KC (S-L2KC) Classifier. Proceedings of 2025 2nd International Conference on Machine Learning and Intelligent Computing, in Proceedings of Machine Learning Research 278:519-526 Available from https://proceedings.mlr.press/v278/lv25a.html.

Related Material