Xinxing Xu (徐新兴)
Artificial Intelligence GroupComputing Science Department
Institute of High Performance Computing (IHPC)
The Agency for Science, Technology and Research (A*STAR)
1 Fusionopolis Way, #16-16 Connexis, 138632, Singapore
Email: xuxi0006 AT ntu DOT edu DOT sg
         xxxing1987 AT gmail DOT com

Xinxing Xu is a scientist with Institute of High Performance Computing (IHPC), Astar, Singapore. He obtained the bachelor's degree in Electronic Engineering and Information Science (EEIS) in University of Science and Technology of China (USTC) in 2009, and also his PhD degree from the Nanyang Technological University (NTU).  His supervisor was Professor Dong Xu, and he also works closely with Associate Processor Ivor Wai-Hung Tsang

2014-5: LibMKL: easy to use matlab code for the soft margin MKLs (including SM1MKL and LpMKL) released!

Research Interests:
  • Machine Learning: Multiple Kernel Learning (MKL), Learning using Privileged Information, Distance Metric Learning
  •                               Deep Learning, Convolution Neural Networks (CNN), Long Short-term Memory (LSTM), Deep Reinforcement Learning
  • Computer Vision: Object Classification, Scene Classification, Video Event Recognition, Video Concept Detection, Action Recognition, Face Verification, Human Gait Recognition, Person Re-identification.
  • Text Categorization: Text Classification, Web Page classification.
  • Recommender System: Click sequence prediction.
Thesis: Learning with Multiple Representations: Algorithms and Applications

  • Shaohua Li, Xinxing Xu, Liqiang Nie, Tat-Seng Chua: Laplacian-Steered Neural Style Transfer. ACM Multimedia (ACM MM), Mountain View, CA USA, Oct 2017. [PDF]
  • Yong Kiam Tan, Xinxing Xu, Yong Liu: Improved Recurrent Neural Networks for Session-based Recommendations. The 1st workshop on Deep Learning for Recommender Systems (DLRS) at the 10th ACM Conference on Recommender Systems (RecSys). Boston, USA, Sept. 2016. [PDF]
  • Xinxing Xu, Joey Tianyi Zhou, IvorW. Tsang, Zheng Qin, Rick Siow Mong Goh, Yong Liu: Simple and Efficient Learning using Privileged Information. BeyondLabeler: Human is More Than a Labeler, Workshop of the 25th International Joint Conference on Artificial Intelligence (IJCAI-16). New York City, USA. July, 2016. [PDF_V1] [PDF_V2]. 2016 Best Paper Award, Contributed Talk.
  • A simple but extremely fast solution for SVM+ by utilizing the squared hinge loss instead of the square loss in the objective function, leading to up to hundred times speed up for SVM+.
  • Joey Tianyi Zhou, Xinxing Xu, Sinno Jialin Pan, Ivor W. Tsang, Qin Zheng and Rick Goh. Transfer Hashing with Privileged Information. In Proceedings of the 25th International Joint Conference on Artificial Intelligence (IJCAI-16). New York City, USA. July 9-15, 2016. [PDF]
  • Li Niu, Xinxing Xu, Lin Chen, Lixin Duan, Dong Xu: Action and Event Recognition in Videos by Learning from Heterogeneous Web Sources. To Appear in IEEE Trans. Neural Netw. Learning Syst. (T-NNLS) [PDF].
  • A novel multi-domain adaptation method based on Elastic-net like Multiple Kernel Learning algorithm by learning from multiple heterogeneous web sources (i.e., Google/Bing Images, Flickr Videos) for video action (e.g., Hollywood2) and event recognition (e.g., CCV).
  • Xinxing Xu, Wen Li, Dong Xu, Ivor W. Tsang: Co-Labeling for Multi-view Weakly Labeled Learning. IEEE Trans. Pattern Anal. Mach. Intell. 38(6): 1113-1125 (2016).  (T-PAMI) [PDF]
  • A novel unified weakly labelled multi-view learning framework using Multi-layer (i.e., 2-layer and 3-layer) Multiple Kernel Learning for Multi-view Semi-supervised learning, Multi-view Multi-instance learning and Multi-view relative outlier detection.
  • Xinxing Xu, Wen Li, Dong Xu: Distance Metric Learning Using Privileged Information for Face Verification and Person Re-Identification.  IEEE Trans. Neural Netw. Learning Syst. 26(12): 3150-3162 (2015). (T-NNLS).
  • A novel Information-theoretic Metric Learning using Privileged Information (ITML+) algorithm for RGB face verification and person re-identification by learning from RGB-D data with additional privileged depth information in the training set.
  • Shengye Yan, Xinxing Xu, Dong Xu, Stephen Lin, and Xuelong Li: Image Classification with Densely Sampled Image Windows and Generalized Adaptive Multiple Kernel Learning. IEEE Transactions on Cybernetics, 45(3): 395-404 (2015). (T-CYB)
  • A Generalized Adaptive Multiple Kernel Learning algorithm by fusing visual features as well as decision features.
  • Shengye Yan, Xinxing Xu, Qingshan Liu: Learning the object location, scale and view for image categorization with adapted classifier. Information Sciences. March, 2014.
  • Xinxing Xu, Ivor W. Tsang and Dong Xu: Soft Margin Multiple Kernel Learning. IEEE Trans. Neural Netw. Learning Syst., vol. 24, no. 5, pp. 749–761, 2013. (T-NNLS) [PDF] [Code]. A new box constrained Multiple Kernel Learning objective and algorithm.
  • A novel Soft Margin framework for Multiple Kernel Learning, unifying, explaining the existing MKL formulations from the margin loss perspective and devising a new regularizer for MKL.
  • Xinxing Xu, Ivor W. Tsang and Dong Xu: Handling Ambiguity via Input-Output Kernel Learning, IEEE Int. Conf. on Data Mining (ICDM), December 2012, pp. 725-734. [PDF] (Full paper, Acceptance rate = 11%)
  • A unified Input-Output Kernel Learning (IOKL) framework based on Group Sparse (2-layer) Multiple Kernel Learning for Multiple Kernel Multiple Instance Learning, Multiple Kernel Semi-supervise Learning, Multiple Kernel Maximum Margin Clustering.
  • Shengye Yan, Xinxing Xu, Dong Xu, Stephen Lin and Xuelong Li: Beyond Spatial Pyramids: A New Feature Extraction Framework with Dense Spatial Sampling for Image Classification, European Conference on Computer Vision (ECCV), 2012
  • [PDF] Multiple Kernel Learning for Caltech256 and 15Scenes.
  • Dong Xu, Yi Huang, Zinan Zeng, Xinxing Xu: Human Gait Recognition Using Patch Distribution Feature and Locality-Constrained Group Sparse Representation. IEEE Transactions on Image Processing (T-IP) 21(1): 316-326 (2012). [PDF]
  • A new patch distribution feature (i.e., referred to as Gabor-PDF), which concatenates the Gabor features together with the X-Y coordinates
    A new classification method called locality-constrained group sparse representation (LGSR), which incoporates both the group sparsity and local smoothness for sparse representation
  • Xinxing Xu, Dong Xu, Ivor W. Tsang: Video Concept Detection Using Support Vector Machine with Augmented Features. Proceedings of the Fourth Pacific-Rim Symposium on Image and Video Technology (PSIVT2010). [PDF]
  • An extremely simple but effective AFSVM, which transfers the knowledge across different semantic classes.


Useful Links:
                             Machine Learning Tutorials               
Deep Reinforcement Learning (David Silver)
                             Deep Reinforcement Learning Papers
Computer Vision: Object Detection

         Huazhu Fu Liang Feng Lin Chen Lixin Duan Shenghua Gao Shijie Xiao Shengye Yan Wen Li

Locations of visitors to this page