Article
Computer Science, Artificial Intelligence
Debamita Kumar, Pradipta Maji
Summary: Multimodal data analysis is used for identifying sample categories by incorporating supervised information and capturing the nonlinear correlated structures across different views. The proposed architecture, discriminative deep canonical correlation analysis (D2CCA), combines generative models with the learning objective to improve discriminative ability. The joint representation of multi-view data is learned from maximally correlated subspaces.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Yinsong Wang, Shahin Shahrampour
Summary: This article proposes a task-specific scoring rule for selecting random features, which can be applied to different applications. The proposed ORCCA method improves upon other approximation techniques in the CCA task by optimizing the scoring function.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Automation & Control Systems
Luefeng Chen, Kuanlin Wang, Min Li, Min Wu, Witold Pedrycz, Kaoru Hirota
Summary: This article proposes a K-means clustering-based kernel canonical correlation analysis algorithm for multimodal emotion recognition in human-robot interaction. By fusing multimodal features from different modalities, the proposed method improves heterogeneity among modalities and enhances the accuracy of emotion recognition.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2023)
Article
Computer Science, Artificial Intelligence
Xianchao Xiu, Lili Pan, Ying Yang, Wanquan Liu
Summary: This study proposes a new joint sparse constrained CCA model that integrates l(2,0)-norm joint sparse constraints into classical CCA for improved fault detection performance. The proposed approach fully exploits the joint sparse structure to determine the number of extracted variables and utilizes an efficient algorithm for computation. Extensive numerical studies demonstrate the efficiency and speed of the proposed method.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Rencheng Song, Jiji Li, Juan Cheng, Chang Li, Yu Liu, Xun Chen
Summary: This article proposes a novel method called robust iBCG (RiBCG) to suppress motion artifacts and an improved version RiBCG-C to reduce HR outliers. Through evaluation on public databases, RiBCG-C method achieves overall the best performance, providing a promising scheme for RiBCG measurements in realistic application scenarios.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Engineering, Electrical & Electronic
Jianqin Sun, Xianchao Xiu, Ziyan Luo, Wanquan Liu
Summary: This study introduces a novel tensor CCA method (TCCA-O) to preserve orthogonality and improve feature representation. By incorporating a structured sparse regularization term (TCCA-OS), the performance of the method is further improved. Experimental results demonstrate that TCCA-O and TCCA-OS outperform other CCA methods in various evaluation metrics.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2023)
Article
Automation & Control Systems
Hongchao Cheng, Jing Wu, Daoping Huang, Yiqi Liu, Qilin Wang
Summary: A novel method called Rab-CCA is proposed for monitoring wastewater treatment processes, which includes a robust decomposition method and an adaptive statistical control limit to improve the performance of standard process monitoring methods, reducing missed alarms and false alarms simultaneously.
Article
Biotechnology & Applied Microbiology
Zishan Wang, Ruqiang Huang, Ye Yan, Zhiguo Luo, Shaokai Zhao, Bei Wang, Jing Jin, Liang Xie, Erwei Yin
Summary: Our study proposed a new method for emotion recognition by exploring the interactions between EEG rhythms under different emotional expressions. The effectiveness of more correlated EEG frequency band features for emotion classification accuracy was verified through experiments. By fusing these correlated features with traditional features at the decision level, we significantly improved the accuracy of emotion recognition.
BIOENGINEERING-BASEL
(2023)
Article
Chemistry, Analytical
Victor Javier Kartsch, Velu Prabhakar Kumaravel, Simone Benatti, Giorgio Vallortigara, Luca Benini, Elisabetta Farella, Marco Buiatti
Summary: Recent studies have shown that the integrity of core perceptual and cognitive functions can be tested with low stimulation frequencies using Steady-State Visual Evoked Potentials (SSVEP) and wearable EEG systems. The results demonstrate that Normalized Canonical Correlation Analysis (NCCA) is an effective method for rapid and accurate detection of SSVEP without the need for preliminary artifact correction or channel selection.
Article
Computer Science, Information Systems
Kai-fa Hui, Ernest Domanaanmwi Ganaa, Yong-zhao Zhan, Xiang-jun Shen
Summary: The paper proposes a method of robust deflated canonical correlation analysis via feature factoring for multi-view image classification, which introduces a feature factoring matrix to evaluate the contribution of each feature to the whole feature space and assign specific weights to different features to suppress noisy data. Multiple factoring matrices are built with respect to multiple projection vectors to weigh the degree of importance of each feature in each projection for better feature representation in multi-view images.
MULTIMEDIA TOOLS AND APPLICATIONS
(2021)
Article
Automation & Control Systems
Xianchao Xiu, Zhonghua Miao, Ying Yang, Wanquan Liu
Summary: This article proposes an efficient nonlinear process monitoring method by integrating DAENNs, CCA, and SCO. The method is demonstrated on the TE process and the diesel generator process, achieving an increased fault detection rate of 8.00% for the fault IDV(11) compared to classical CCA.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2022)
Article
Mathematics, Applied
Gang Wu, Fei Li
Summary: Canonical correlation analysis (CCA) is a widely used data analysis method that aims to extract meaningful information by maximizing the correlation between two sets of variables. However, it may face challenges such as small sample size and overfitting. In this work, an Exponential Canonical Correlation Analysis (ECCA) method based on matrix exponential is proposed, as well as a Randomized Exponential Canonical Correlation Analysis (RECCA) method using randomized singular value decomposition to address these challenges. Experimental results demonstrate the superior performance of these proposed algorithms over existing CCA methods.
APPLIED NUMERICAL MATHEMATICS
(2021)
Article
Computer Science, Artificial Intelligence
Xiang-Jun Shen, Zhaorui Xu, Liangjun Wang, Zechao Li, Guangcan Liu, Jianping Fan, ZhengJun Zha
Summary: This article presents a novel CCA method that carries out analysis on the dataset in the Fourier domain, which can significantly improve the computation speed and memory efficiency. By applying Fourier transform on the data, the traditional eigenvector computation of CCA is converted into finding discriminative Fourier bases. The proposed method achieves satisfactory accuracy and extremely fast training time consumption in large-scale correlation datasets.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Automation & Control Systems
Hongtian Chen, Zhiwen Chen, Zheng Chai, Bin Jiang, Biao Huang
Summary: This study introduces a new nonlinear fault detection method called SsCCA with the help of neural networks to enhance FD performance, by reformulating the objective function and designing a specific solution. Experimental results demonstrate that this method can effectively improve fault detection capability in nonlinear systems.
IEEE TRANSACTIONS ON CYBERNETICS
(2022)
Article
Computer Science, Artificial Intelligence
Paris A. Karakasis, Athanasios P. Liavas, Nicholas D. Sidiropoulos, Panagiotis G. Simos, Efrosini Papadaki
Summary: Functional magnetic resonance imaging (fMRI) is widely used for studying the human brain. This study proposes a new fMRI data generating model that takes into account both task-related and resting-state components. Experimental tests show that our method can accurately estimate temporal and spatial components even at low Signal to Noise Ratio (SNR), and outperforms standard procedures based on General Linear Models (GLMs).
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2022)
Article
Statistics & Probability
Tamara Fernandez, Arthur Gretton, David Rindt, Dino Sejdinovic
Summary: The method introduces a general nonparametric independence test, which can effectively detect complex nonlinear dependencies. It is computationally straightforward and outperforms competing methods.
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
(2023)
Article
Computer Science, Artificial Intelligence
Vignesh Srinivasan, Klaus-Robert Mueller, Wojciech Samek, Shinichi Nakajima
Summary: In this article, the authors propose a strategy called Langevin cooling (L-Cool) to enhance existing methods in image translation and language translation tasks. They suggest using Langevin dynamics to bring fringe samples from low-density areas to high-density areas.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Biochemistry & Molecular Biology
Philipp Keyl, Philip Bischoff, Gabriel Dernbach, Michael Bockmayr, Rebecca Fritz, David Horst, Nils Bluethgen, Gregoire Montavon, Klaus-Robert Mueller, Frederick Klauschen
Summary: The molecular heterogeneity of cancer cells contributes to the often partial response to targeted therapies and relapse of disease due to the escape of resistant cell populations. Single-cell sequencing has limitations in understanding this heterogeneity, and scGeneRAI is an explainable deep learning approach that can infer gene regulatory networks from single-cell RNA sequencing data to provide functional insights. Our method reveals characteristic network patterns for tumor cells and normal epithelial cells and allows the reconstruction of networks at the level of single cells, which helps characterize the heterogeneity of gene regulation within and across tumors.
NUCLEIC ACIDS RESEARCH
(2023)
Article
Multidisciplinary Sciences
Stefan Chmiela, Valentin Vassilev-Galindo, Oliver T. Unke, Adil Kabylda, Huziel E. Sauceda, Alexandre Tkatchenko, Klaus-Robert Mueller
Summary: We have developed an exact iterative approach to train global symmetric gradient domain machine learning (sGDML) force fields, which can accurately describe complex molecular systems and materials. We evaluated the accuracy and efficiency of sGDML on a newly developed MD22 benchmark dataset containing molecules from 42 to 370 atoms.
Article
Chemistry, Physical
Stefan Bluecher, Klaus-Robert Mueller, Stefan Chmiela
Summary: Kernel machines have achieved continuous progress in the field of quantum chemistry, especially in the low-data regime of force field reconstruction. However, the scalability of kernel machines has been hindered by their quadratic memory and cubical runtime complexity.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Biochemical Research Methods
Kai J. Miller, Klaus-Robert J. Mueller, Gabriela Ojeda Valencia, Harvey J. Huang, Nicholas M. Gregg, Gregory A. J. Worrell, Dora Hermes
Summary: Single-pulse electrical stimulation, known as cortico-cortical evoked potential measurement, provides a valuable technique for understanding brain region interactions. This study presents a new method, called Canonical Response Parameterization (CRP), to quantify brain stimulation data and compare voltage response traces from different areas. This technique enables the quantification of various parameters, such as cross-projection magnitudes, response duration, canonical shape projection amplitudes, signal-to-noise ratios, explained variance, and statistical significance.
PLOS COMPUTATIONAL BIOLOGY
(2023)
Article
Mathematics
Alexander Bauer, Shinichi Nakajima, Klaus-Robert Mueller
Summary: The paper introduces an efficient exact inference method for local models, which allows for finer interactions between the energy of the core model and the sufficient statistics of the global terms. This greatly increases the range of admissible applications and improves upon the theoretical guarantees of computational efficiency.
Article
Clinical Neurology
Katrin S. Wendrich, Hamid Azimi, Jurgen A. Ripperger, Yann Ravussin, Gregor Rainer, Urs Albrecht
Summary: The sleep-wake cycle is regulated by the circadian clock and the sleep homeostat. It is not completely understood how these two systems interact. Evidence suggests that the clock gene Per2 may be involved in the sleep homeostatic process. Neurons and astroglial cells in the brain depend on each other metabolically and play a role in sleep regulation. The study found that mice lacking Per2 in all body cells displayed earlier sleep onset after sleep deprivation, while mice lacking Per2 in neurons or astroglial cells were normal in this regard. Systemic Per2 expression seems to be important for the sleep architecture, while neuronal and astroglial Per2 weakly affects sleep amount. The results indicate that Per2 contributes to the timing of the sleep response by delaying sleep onset after sleep deprivation and attenuating the early night rebound response.
Article
Computer Science, Artificial Intelligence
Lorenz Linhardt, Klaus-Robert Mueller, Gregoire Montavon
Summary: This paper investigates the issue of mismatches between the decision strategy of the explainable model and the user's domain knowledge, and proposes a new method EGEM to mitigate hidden flaws in the model. Experimental results demonstrate that the approach can significantly reduce reliance on Clever Hans strategies and improve the accuracy of the model on new data.
INFORMATION FUSION
(2024)
Article
Chemistry, Physical
Joshua Scheidt, Alexander Diener, Michael Maiworm, Klaus-Robert Mueller, Rolf Findeisen, Kurt Driessens, F. Stefan Tautz, Christian Wagner
Summary: A nanofabrication technique involving the assembly of functional molecular structures using a scanning probe microscope (SPM) has been developed. The key challenge was the lack of simultaneous actuation and imaging capabilities of the SPM tip, which hindered continuous monitoring of molecular configuration during manipulation. However, in this study, configuration monitoring was achieved through modelling the manipulation process as a partially observable Markov decision process (POMDP) and using a particle filter. The proposed methodology is an important step towards the robotic and possibly automated creation of supramolecular structures.
JOURNAL OF PHYSICAL CHEMISTRY C
(2023)
Article
Chemistry, Physical
Bipeng Wang, Ludwig Winkler, Yifan Wu, Klaus-Robert Muller, Huziel E. Sauceda, Oleg V. Prezhdo
Summary: Nonadiabatic molecular dynamics (MD) is crucial for understanding far-from-equilibrium processes, but requires expensive calculations of excitation energies and nonadiabatic couplings. In this study, a bidirectional long short-term memory network (Bi-LSTM) is employed in the time domain to interpolate the Hamiltonian, achieving significant computational savings compared to direct ab initio calculation. The Bi-LSTM-NAMD method outperforms previous models and captures slow and fast time scales, extending MD simulation times from picoseconds to nanoseconds.
JOURNAL OF PHYSICAL CHEMISTRY LETTERS
(2023)
Article
Automation & Control Systems
Armin W. Thomas, Ulman Lindenberger, Wojciech Samek, Klaus-Robert Mueller
Summary: Research has shown that transfer learning improves the performance of deep learning models in datasets with small sample sizes. In this study, the application of transfer learning to cognitive decoding analysis using functional neuroimaging data is systematically evaluated. Pre-trained deep learning models consistently achieve higher decoding accuracies and require less training time and data compared to models trained from scratch. The benefits of pre-training come from the ability to reuse learned features when training with new data. However, challenges arise when interpreting the decoding decisions of pre-trained models, as they may utilize fMRI data in unforeseen and counterintuitive ways.
JOURNAL OF THE FRANKLIN INSTITUTE-ENGINEERING AND APPLIED MATHEMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Leo Andeol, Yusei Kawakami, Yuichiro Wada, Takafumi Kanamori, Klaus-Robert Mueller, Gregoire Montavon
Summary: This paper addresses the domain shift problem in machine learning and proposes a method that can generate more invariant representations and more stable prediction performance across different domains, using mathematical relations with the Wasserstein distance. Empirical results on multiple image datasets show the effectiveness of the proposed approach.
Article
Psychology, Educational
Amna Ghani, Caroline Di Bernardi Luft, Smadar Ovadio-Caro, Klaus-Robert Mueller, Joydeep Bhattacharya
Summary: Chance favors the prepared mind, said Louis Pasteur. In this study, the researchers investigated the brain's receptivity to integrate new information and the experience of creative insights known as Aha! moments. They hypothesized that the transient oscillatory states of the brain would characterize its preparedness for these insights. Through a real-time brain-state-dependent cognitive stimulation experiment, they found that participants were more successful in utilizing clues and experienced more Aha responses when clues were presented at the up-regulated state of right temporal alpha oscillation. Additionally, they observed a negative correlation between the coupling of alpha oscillation phase and gamma oscillation power and the frequency of Aha moments. These findings highlight the role of brain oscillations in the Aha experience and provide insights into the neural mechanism underlying the brain's receptivity to integrate semantic information.
CREATIVITY RESEARCH JOURNAL
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Andreas Holzinger, Randy Goebel, Ruth Fong, Taesup Moon, Klaus-Robert Mueller, Wojciech Samek
Summary: Explainable Artificial Intelligence (XAI) is a research field focused on explaining the decisions and internal mechanisms of complex AI systems. It aims to provide methods to interpret and explain the results of AI methods, particularly those that are difficult for human experts to comprehend. XAI is crucial for establishing trust in AI applications that impact human life and can be applied to various types of models using diverse array of techniques.
XXAI - BEYOND EXPLAINABLE AI: International Workshop, Held in Conjunction with ICML 2020, July 18, 2020, Vienna, Austria, Revised and Extended Papers
(2022)