1 code implementation • 8 Dec 2023 • Xiaoyun Xu, Shujian Yu, Jingzheng Wu, Stjepan Picek
However, these methods still follow the design of traditional supervised adversarial training, limiting the potential of adversarial training on ViTs.
no code implementations • 26 Oct 2023 • Qiang Li, Vince D. Calhoun, Adithya Ram Ballem, Shujian Yu, Jesus Malo, Armin Iraji
The human brain has a complex, intricate functional architecture.
no code implementations • 21 Oct 2023 • Francesco Alesiani, Shujian Yu, Mathias Niepert
Invariant risk minimization (IRM) is a recent proposal for discovering environment-invariant representations.
1 code implementation • 17 Jul 2023 • Luis Pedro Silvestrin, Shujian Yu, Mark Hoogendoorn
In this paper, we revisit the robustness of the minimum error entropy (MEE) criterion, a widely used objective in statistical signal processing to deal with non-Gaussian noises, and investigate its feasibility and usefulness in real-life transfer learning regression tasks, where distributional shifts are common.
1 code implementation • 8 Jun 2023 • Shuo Ye, Shujian Yu, Wenjin Hou, Yu Wang, Xinge You
Fine-grained visual categorization (FGVC) is a challenging task due to similar visual appearances between various species.
no code implementations • 21 Mar 2023 • Qiang Li, Shujian Yu, Kristoffer H Madsen, Vince D Calhoun, Armin Iraji
To address this problem, we applied multivariate mutual information, specifically, Total Correlation and Dual Total Correlation to reveal higher-order information in the brain.
no code implementations • 21 Jan 2023 • Shujian Yu, Hongming Li, Sigurd Løkse, Robert Jenssen, José C. Príncipe
In this paper, we extend the classic CS divergence to quantify the closeness between two conditional distributions and show that the developed conditional CS divergence can be simply estimated by a kernel density estimator from given samples.
1 code implementation • 16 Jan 2023 • Hongming Li, Shujian Yu, Jose Principe
We propose causal recurrent variational autoencoder (CR-VAE), a novel generative model that is able to learn a Granger causal graph from a multivariate time series x and incorporates the underlying causal mechanism into its data generation process.
1 code implementation • 4 Jan 2023 • Kaizhong Zheng, Shujian Yu, Badong Chen
There is a recent trend to leverage the power of graph neural networks (GNNs) for brain-network based psychiatric diagnosis, which, in turn, also motivates an urgent need for psychiatrists to fully understand the decision behavior of the used GNNs.
1 code implementation • 30 Nov 2022 • Yuxin Dong, Tieliang Gong, Shujian Yu, Hong Chen, Chen Li
The matrix-based R\'enyi's entropy allows us to directly quantify information measures from given data, without explicit estimation of the underlying probability distribution.
no code implementations • 6 Oct 2022 • Qiang Li, Greg Ver Steeg, Shujian Yu, Jesus Malo
In this work we build on this idea to infer a large scale (whole brain) connectivity network based on Total Correlation and show the possibility of using this kind of networks as biomarkers of brain alterations.
no code implementations • 26 Sep 2022 • Yufeng Shi, Shujian Yu, Duanquan Xu, Xinge You
In this paper, instead of using an extra NLP model to define a common space beforehand, we consider a totally different way to construct (or learn) a common hamming space from an information-theoretic perspective.
1 code implementation • 31 May 2022 • Shujian Yu, Francesco Alesiani, Wenzhe Yin, Robert Jenssen, Jose C. Principe
Graph sparsification aims to reduce the number of edges of a graph while maintaining its structural properties.
no code implementations • 16 May 2022 • Yuxin Dong, Tieliang Gong, Shujian Yu, Chen Li
The Matrix-based Renyi's entropy enables us to directly measure information quantities from given data without the costly probability density estimation of underlying distributions, thus has been widely adopted in numerous statistical learning and inference tasks.
no code implementations • 7 May 2022 • Kaizhong Zheng, Shujian Yu, Baojuan Li, Robert Jenssen, Badong Chen
Developing a new diagnostic models based on the underlying biological mechanisms rather than subjective symptoms for psychiatric disorders is an emerging consensus.
1 code implementation • 22 Apr 2022 • Qi Zhang, Shujian Yu, Jingmin Xin, Badong Chen
By "intelligently" fusing the complementary information across different views, multi-view learning is able to improve the performance of classification tasks.
no code implementations • 21 Apr 2022 • Yu Wang, Shuo Ye, Shujian Yu, Xinge You
In this paper, we present a novel approach for FGVC, which can simultaneously make use of partial yet sufficient discriminative information in environmental cues and also compress the redundant information in class-token with respect to the target.
1 code implementation • 7 Feb 2022 • Hongming Li, Shujian Yu, Jose C. Principe
We develop a new neural network based independent component analysis (ICA) method by directly minimizing the dependence amongst all extracted components.
no code implementations • 27 Dec 2021 • Tieliang Gong, Yuxin Dong, Shujian Yu, Bo Dong
The recently developed matrix based Renyi's entropy enables measurement of information in data simply using the eigenspectrum of symmetric positive semi definite (PSD) matrices in reproducing kernel Hilbert space, without estimation of the underlying data distribution.
1 code implementation • 12 Oct 2021 • Bo Hu, Shujian Yu, Jose C. Principe
We test the framework for estimation of mutual information and compare the results with the mutual information neural estimation (MINE), for density estimation, for conditional probability estimation in Markov models as well as for training adversarial networks.
1 code implementation • 12 Oct 2021 • Francesco Alesiani, Shujian Yu, Xi Yu
By learning minimum sufficient representations from training data, the information bottleneck (IB) approach has demonstrated its effectiveness to improve generalization in different AI applications.
Adversarial Robustness Out of Distribution (OOD) Detection +1
no code implementations • 7 Aug 2021 • Ammar Shaker, Shujian Yu, Daniel Oñoro-Rubio
Feature similarity includes both the invariance of marginal distributions and the closeness of conditional distributions given the desired response $y$ (e. g., class labels).
1 code implementation • 31 Jan 2021 • Xi Yu, Shujian Yu, Jose C. Principe
We introduce the matrix-based Renyi's $\alpha$-order entropy functional to parameterize Tishby et al. information bottleneck (IB) principle with a neural network.
1 code implementation • 25 Jan 2021 • Shujian Yu, Francesco Alesiani, Xi Yu, Robert Jenssen, Jose C. Principe
Measuring the dependence of data plays a central role in statistics and machine learning.
no code implementations • 2 Nov 2020 • Ammar Shaker, Shujian Yu, Francesco Alesiani
In this paper, we propose a continual learning (CL) technique that is beneficial to sequential task learners by improving their retained accuracy and reducing catastrophic forgetting.
no code implementations • 2 Nov 2020 • Ammar Shaker, Francesco Alesiani, Shujian Yu, Wenzhe Yin
This paper presents Bilevel Continual Learning (BiCL), a general framework for continual learning that fuses bilevel optimization and recent advances in meta-learning for deep neural networks.
no code implementations • 11 Sep 2020 • Francesco Alesiani, Shujian Yu, Ammar Shaker, Wenzhe Yin
Interpretable Multi-Task Learning can be expressed as learning a sparse graph of the task relationship based on the prediction performance of the learned models.
no code implementations • 11 Sep 2020 • Shujian Yu, Francesco Alesiani, Ammar Shaker, Wenzhe Yin
We present a novel methodology to jointly perform multi-task learning and infer intrinsic relationship among tasks by an interpretable and sparse graph.
1 code implementation • 21 Jul 2020 • Feiya Lv, Shujian Yu, Chenglin Wen, Jose C. Principe
This paper presents a novel mutual information (MI) matrix based method for fault detection.
no code implementations • 13 Jul 2020 • Yanjun Li, Shujian Yu, Jose C. Principe, Xiaolin Li, Dapeng Wu
Although substantial efforts have been made to learn disentangled representations under the variational autoencoder (VAE) framework, the fundamental properties to the dynamics of learning of most VAE models still remain unknown and under-investigated.
1 code implementation • 12 May 2020 • Shiyu Duan, Shujian Yu, Jose Principe
By redefining the conventional notions of layers, we present an alternative view on finitely wide, fully trainable deep neural networks as stacked linear models in feature spaces, leading to a kernel machine interpretation.
1 code implementation • 5 May 2020 • Shujian Yu, Ammar Shaker, Francesco Alesiani, Jose C. Principe
We propose a simple yet powerful test statistic to quantify the discrepancy between two conditional distributions.
no code implementations • 25 Sep 2019 • Kristoffer Wickstrøm, Sigurd Løkse, Michael Kampffmeyer, Shujian Yu, Jose Principe, Robert Jenssen
In this paper, we propose an IP analysis using the new matrix--based R\'enyi's entropy coupled with tensor kernels over convolutional layers, leveraging the power of kernel methods to represent properties of the probability distribution independently of the dimensionality of the data.
no code implementations • 25 Sep 2019 • Kristoffer Wickstrøm, Sigurd Løkse, Michael Kampffmeyer, Shujian Yu, Jose Principe, Robert Jenssen
In this paper, we propose an IP analysis using the new matrix--based R\'enyi's entropy coupled with tensor kernels over convolutional layers, leveraging the power of kernel methods to represent properties of the probability distribution independently of the dimensionality of the data.
1 code implementation • 13 Jul 2019 • Yantao Wei, Shujian Yu, Luis Sanchez Giraldo, Jose C. Principe
This paper proposes a novel architecture, termed multiscale principle of relevant information (MPRI), to learn discriminative spectral-spatial features for hyperspectral image (HSI) classification.
no code implementations • 29 May 2019 • Zhengqiang Zhang, Shujian Yu, Shi Yin, Qinmu Peng, Xinge You
Weakly-supervised semantic segmentation aims to assign each pixel a semantic category under weak supervisions, such as image-level tags.
no code implementations • ICLR 2019 • Shiyu Duan, Shujian Yu, Yun-Mei Chen, Jose Principe
Moreover, unlike backpropagation, which turns models into black boxes, the optimal hidden representation enjoys an intuitive geometric interpretation, making the dynamics of learning in a deep kernel network simple to understand.
no code implementations • 29 Nov 2018 • Shujian Yu, Jose C. Principe
Feature selection aims to select the smallest feature subset that yields the minimum generalization error.
1 code implementation • 19 Nov 2018 • Peng Zhang, Shujian Yu, Jiamiao Xu, Xinge You, Xiubao Jiang, Xiao-Yuan Jing, DaCheng Tao
It remains a huge challenge to design effective and efficient trackers under complex scenarios, including occlusions, illumination changes and pose variations.
1 code implementation • 23 Aug 2018 • Shujian Yu, Luis Gonzalo Sanchez Giraldo, Robert Jenssen, Jose C. Principe
The matrix-based Renyi's \alpha-order entropy functional was recently introduced using the normalized eigenspectrum of a Hermitian matrix of the projected data in a reproducing kernel Hilbert space (RKHS).
no code implementations • 25 Jun 2018 • Shujian Yu, Xiaoyang Wang, Jose C. Principe
In this paper, a novel Hierarchical Hypothesis Testing framework with Request-and-Reverify strategy is developed to detect concept drifts by requesting labels only when necessary.
no code implementations • 21 May 2018 • Qi Zheng, Shujian Yu, Xinge You, Qinmu Peng
Low-Rank Matrix Recovery (LRMR) has recently been applied to saliency detection by decomposing image features into a low-rank component associated with background and a sparse component associated with visual salient regions.
no code implementations • 19 Apr 2018 • Jiamiao Xu, Shujian Yu, Xinge You, Mengjun Leng, Xiao-Yuan Jing, C. L. Philip Chen
We present a novel cross-view classification algorithm where the gallery and probe data come from different views.
no code implementations • 18 Apr 2018 • Shujian Yu, Kristoffer Wickstrøm, Robert Jenssen, Jose C. Principe
The matrix-based Renyi's \alpha-entropy functional and its multivariate extension were recently developed in terms of the normalized eigenspectrum of a Hermitian matrix of the projected data in a reproducing kernel Hilbert space (RKHS).
no code implementations • 30 Mar 2018 • Shujian Yu, Jose C. Principe
Despite their great success in practical applications, there is still a lack of theoretical and systematic methods to analyze deep neural networks.
1 code implementation • ICLR 2019 • Shiyu Duan, Shujian Yu, Yun-Mei Chen, Jose Principe
With this method, we obtain a counterpart of any given NN that is powered by kernel machines instead of neurons.
no code implementations • 25 Jul 2017 • Shujian Yu, Zubin Abraham, Heng Wang, Mohak Shah, Yantao Wei, José C. Príncipe
A fundamental issue for statistical classification models in a streaming environment is that the joint distribution between predictor and response variables changes over time (a phenomenon also known as concept drifts), such that their classification performance deteriorates dramatically.
no code implementations • 3 May 2017 • Zheng Cao, Shujian Yu, Bing Ouyang, Fraser Dalgleish, Anni Vuorenkoski, Gabriel Alsenas, Jose Principe
Depending on the quantity and properties of acquired imagery, the animals are characterized as either features (shape, color, texture, etc.