Article
Computer Science, Artificial Intelligence
Nicolas Nadisic, Jeremy E. Cohen, Arnaud Vandaele, Nicolas Gillis
Summary: This paper introduces a new form of sparse MNNLS problem and a two-step algorithm to solve it. By dividing the problem into subproblems and selecting Pareto front solutions, a matrix that satisfies the sparsity constraint is constructed. Experimental results show that this method is more accurate than existing heuristic algorithms.
Article
Mathematics, Applied
Monica Dessole, Marco Dell'Orto, Fabio Marcuzzi
Summary: This paper presents an improved version of the LHDM method, which can terminate in a finite number of steps and is applicable to a wider range of matrix categories compared to the previous version. In addition, when solving underdetermined linear systems using the NNLS method, LHDM can find sparser solutions. Extensive experiments are conducted to evaluate the performance improvement of LHDM compared to the standard Lawson-Hanson algorithm, and it is compared to several l1-minimization solvers in terms of solution quality and time-to-solution on a large set of dense instances.
NUMERICAL LINEAR ALGEBRA WITH APPLICATIONS
(2023)
Article
Engineering, Electrical & Electronic
Bardia Yousefi, Clemente Ibarra Castanedo, Xavier P. V. Maldague
Summary: This study conducts a comparative analysis on low-rank matrix approximation methods in thermography and demonstrates the practicality and efficiency of semi-, convex-, and sparse-nonnegative matrix factorization methods for detecting subsurface thermal patterns. The experimental results show that these methods are effective in subsurface defect detection and distinguishing breast abnormalities in breast cancer screening data sets.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2021)
Article
Geochemistry & Geophysics
Le Dong, Yuan Yuan, Xiaoqiang Lu
Summary: In this article, a novel unmixing method is proposed using two types of self-similarity to constrain sparse NMF. The method explores spatial global self-similarity groups between pixels based on the whole image and creates spectral local self-similarity groups inside superpixels. By sparsely encoding pixels within each spatial and spectral group, the method forces the abundance of pixels within each group to be similar, thereby constraining the NMF unmixing framework. Experiments demonstrate the superiority of this method over existing methods on synthetic and real data.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2021)
Article
Computer Science, Interdisciplinary Applications
M. Karimpour, M. Rezghi
Summary: This paper introduces an efficient algorithm for NMF decomposition based on the ANLS framework, using the BCI-NC method to solve nonnegativity constrained least-squares subproblems. Experimental results show that the algorithm performs efficiently on image and text datasets.
JOURNAL OF COMPUTATIONAL SCIENCE
(2022)
Article
Geochemistry & Geophysics
Taner Ince, Nicolas Dobigeon
Summary: A weighted residual nonnegative matrix factorization method with spatial regularization is proposed to unmix hyperspectral data, aiming to improve the robustness of NMF against noise. Experimental results validate the effectiveness of the proposed method in providing spatial information for abundance matrix.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Mathematics, Applied
N. Waniorek, D. Calvetti, E. Somersalo
Summary: Dictionary learning has been widely used in various applications, such as image denoising, face recognition, remote sensing, medical imaging, and feature extraction, to represent a signal using dictionary atoms. Sparse dictionary learning is particularly interesting when the signal can be represented by a few vectors in a given basis. This paper proposes using hierarchical Bayesian models for sparse dictionary learning, which can capture features of the underlying signals, such as sparse representation and nonnegativity. The framework can also be used for dimensionality reduction of an annotated dictionary through feature extraction, reducing the computational complexity of the learning task.
Article
Mathematical & Computational Biology
Weijuan Liang, Shuangge Ma, Qingzhao Zhang, Tingyu Zhu
Summary: Partial least squares is important for dimension reduction in handling problems with numerous variables. Sparse partial least squares technique helps identify important variables and generate more interpretable results. Integrative analysis gathers raw data from multiple independent datasets and jointly analyzes them to improve performance.
STATISTICS IN MEDICINE
(2021)
Article
Computer Science, Artificial Intelligence
Athanasios I. Salamanis, George A. Gravvanis, Sotiris Kotsiantis, Konstantinos M. Giannoutakis
Summary: Although existing missing data imputation methods mainly focus on either time series or tabular data, this paper proposes a generic sparse regression method that can handle missing data in both types of data. The method utilizes a preconditioned iterative approach based on generic approximate sparse pseudoinverse to solve a sparse least squares problem, and introduces sparsity by dummy encoding categorical features. Extensive experiments conducted on multiple datasets demonstrate the effectiveness of the proposed method.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Mathematics, Applied
Jaroslav M. Fowkes, Nicholas I. M. Gould, Jennifer A. Scott
Summary: This article introduces a novel approach for approximate computation of large sparse Hessian matrices. By reformulating the problem as a large linear least squares problem and leveraging recent research on solving such problems, a robust computation method is proposed. The effectiveness and robustness of this method are demonstrated through examples of sparse Hessians from the CUTEst test problem collection for optimization.
NUMERICAL ALGORITHMS
(2023)
Article
Engineering, Electrical & Electronic
Yan-Chong Song, Fei-Yun Wu, Ru Peng
Summary: This article presents the orthogonal least squares (OLS) algorithm and its limitations in improving reconstruction accuracy. A neighborhood-based multiple orthogonal least squares (NMOLS) algorithm is proposed to address this issue.
Article
Mathematics, Applied
Abeynaya Gnanasekaran, Eric Darve
Summary: In this work, a fast hierarchical solver is developed for solving large, sparse least squares problems. The solver utilizes a multifrontal QR approach and low-rank approximation to reduce computational complexity and memory usage, resulting in an approximate factorization of the matrix stored as a sequence of sparse orthogonal and upper-triangular factors.
JOURNAL OF SCIENTIFIC COMPUTING
(2022)
Article
Computer Science, Software Engineering
Srinivas Eswar, Koby Hayashi, Grey Ballard, Ramakrishnan Kannan, Michael A. Matheson, Haesun Park
Summary: This study addresses the problem of low-rank approximation of massive dense nonnegative tensor data by proposing a distributed-memory parallel computing solution that loads input data across multiple nodes and uses efficient and scalable parallel algorithms. The software package presented allows for extension in terms of data, algorithm, and architecture, while carefully avoiding unnecessary communication and computation. Efficiency and scalability results are reported for both synthetic and real-world data sets.
ACM TRANSACTIONS ON MATHEMATICAL SOFTWARE
(2021)
Article
Mathematics, Applied
Xue-Feng Duan, Juan Li, Shan-Qi Diuan, Qing-Wen Wang
Summary: This paper considers the generalized nonnegative tensor factorization (GNTF) problem and proposes a proximal alternating nonnegative least squares method to solve it, along with proving its convergence theorem. Numerical examples demonstrate the feasibility and effectiveness of the new method.
NUMERICAL ALGORITHMS
(2021)
Article
Physics, Multidisciplinary
Jiaming Yu, Hui Qi, Xiangyu Li, Kai Wang, Jing Guo
Summary: Nonlinear aeroelastic systems are difficult to model and calculate due to their complex structure and dynamic response. Model identification is an attractive method for analyzing such systems. However, traditional methods often produce complex models with limited applicability, necessitating the development of interpretable reduced models. This paper proposes a sparse identification method for complex aeroelastic systems using the sparse regression method and sequential threshold least squares technique. The identified models contain only the necessary nonlinear terms based on measurement data. The method is applied to identify a binary wing with dead zone nonlinearity and cubic stiffness nonlinearity, and the resulting model enables rapid and accurate prediction of system response and serves as an explicit surrogate model for aeroelastic optimization design, demonstrating its superiority.
Article
Multidisciplinary Sciences
Simon Reich, Dajie Zhang, Tomas Kulvicius, Sven Boelte, Karin Nielsen-Saines, Florian B. Pokorny, Robert Peharz, Luise Poustka, Florentin Woergoetter, Christa Einspieler, Peter B. Marschik
Summary: This study introduces a novel machine learning algorithm to detect age-specific movement patterns in infants. By analyzing skeletal information from video data, the algorithm is able to accurately discriminate fidgety movements with a classification accuracy of 88%, potentially becoming a universally accessible tool in clinical practice.
SCIENTIFIC REPORTS
(2021)
Article
Computer Science, Artificial Intelligence
Xiaoting Shao, Alejandro Molina, Antonio Vergari, Karl Stelzner, Robert Peharz, Thomas Liebig, Kristian Kersting
Summary: This paper introduces a new type of conditional SPNs model, which can be used as tractable building blocks of deep probabilistic models and shows competitive performance in various tasks. The research demonstrates that CSPNs can outperform other probabilistic models and have the potential to improve performance in different tasks.
INTERNATIONAL JOURNAL OF APPROXIMATE REASONING
(2022)
Article
Acoustics
Lukas Pfeifenberger, Franz Pernkopf
Summary: The paper introduces the BSSD network, which achieves speaker separation, dereverberation, and speaker identification simultaneously. Various techniques like predefined spatial cues, neural beamforming, embedding vectors, and triplet mining are utilized for these tasks. The system is evaluated based on SI-SDR, WER, and EER metrics.
SPEECH COMMUNICATION
(2022)
Article
Engineering, Biomedical
Truc Nguyen, Franz Pernkopf
Summary: Computational methods for lung sound analysis are used for computer-aided diagnosis support, storage, and monitoring in critical care. This paper proposes a method that uses pre-trained ResNet models for the classification of adventitious lung sounds and respiratory diseases. The method incorporates techniques such as fine-tuning, co-tuning, stochastic normalization, and data augmentation to improve performance. The experimental results show that the proposed systems outperform state-of-the-art lung sound classification systems.
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING
(2022)
Article
Physics, Nuclear
Christoph Obermair, Thomas Cartier-Michaud, Andrea Apollonio, William Millar, Lukas Felsberger, Lorenz Fischl, Holger Severin Bovbjerg, Daniel Wollmann, Walter Wuensch, Nuria Catalan-Lasheras, Marca Boronat, Franz Pernkopf, Graeme Burt
Summary: This paper explores the occurrence of vacuum arcs or radio frequency breakdowns in particle accelerators, which limit the high-gradient performance of normal conducting rf cavities. Using a machine learning strategy and explainable artificial intelligence, the authors reverse-engineer physical properties to derive fast and reliable rule-based models. The results show the correlation between the field emitted current following an initial breakdown and the probability of another breakdown occurring shortly thereafter.
PHYSICAL REVIEW ACCELERATORS AND BEAMS
(2022)
Article
Computer Science, Artificial Intelligence
Christian Knoll, Adrian Weller, Franz Pernkopf
Summary: Self-guided belief propagation (SBP) is an enhanced version of belief propagation that gradually incorporates pairwise potentials, increasing accuracy without increasing computational burden. SBP finds the global optimum of the Bethe approximation for attractive models and obtains a unique, stable, and accurate solution when BP does not converge.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Proceedings Paper
Acoustics
Barbara Schuppler, Emil Berger, Xenia Kogler, Franz Pernkopf
Summary: The high degree of segmental reduction in conversational speech leads to a large number of words becoming homophones, which poses a challenge for automatic speech recognition. This study proposes two approaches, one based on prosodic and spectral features and the other based on a convolutional neural network, to disambiguate homophones. The results show potential for both approaches, especially when combined with a stochastic language model as part of an ASR system.
Proceedings Paper
Engineering, Electrical & Electronic
Michael Hirschmugl, Johanna Rock, Paul Meissner, Franz Pernkopf
Summary: In this paper, the authors propose a scalable and resource-efficient accelerator for interference mitigation in automotive radar. They compare different implementations and conclude that hardware acceleration significantly improves speed without much increase in power consumption, and the use of integer arithmetic does not compromise performance. The parallelization of layers in programmable logic further reduces latency.
2022 19TH EUROPEAN RADAR CONFERENCE (EURAD)
(2022)
Proceedings Paper
Acoustics
David Peter, Wolfgang Roth, Franz Pernkopf
Summary: This paper introduces a neural architecture search (NAS) approach for automatically discovering end-to-end keyword spotting (KWS) models in limited resource environments. The authors optimize the structure of convolutional neural networks (CNNs) operating on raw audio waveforms using a differentiable NAS method. After finding a suitable KWS model with NAS, weight and activation quantization is performed to reduce memory usage. Extensive experiments are conducted on the Google speech commands dataset, comparing the end-to-end models to mel-frequency cepstral coefficient (MFCC) based CNNs. Using only NAS, a highly efficient model with an accuracy of 95.55% is obtained, using 75.7k parameters and 13.6M operations. By using trained bit-width quantization, the same model achieves a test accuracy of 93.76% while utilizing on average only 2.91 bits per activation and 2.51 bits per weight.
2022 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP)
(2022)
Proceedings Paper
Audiology & Speech-Language Pathology
Lukas Pfeifenberger, Matthias Zoehrer, Franz Pernkopf
Summary: The paper introduces the CDEC system for the Interspeech 2021 AEC-Challenge, which includes TDC module, frequency-domain AEC, and time-domain neural network, achieving an overall MOS score of 3.80.
Proceedings Paper
Audiology & Speech-Language Pathology
Stefan Fragner, Tobias Topar, Maximilian Giller, Lukas Pfeifenberger, Franz Pernkopf
Summary: This paper introduces an autonomous robot for recording a database of Room Impulse Responses (RIRs) at a high spatial resolution, which can be used to create realistic simulation environments. These RIRs can be exploited to generate multi-channel speech mixtures of static or moving speakers for various applications.
Proceedings Paper
Engineering, Biomedical
Truc Nguyen, Franz Pernkopf
Summary: This study utilizes transfer learning to address the issue of mismatched recording setups in small non-public data, transferring knowledge from one dataset to another for crackle detection in lung sounds and achieving significant performance improvements.
2021 43RD ANNUAL INTERNATIONAL CONFERENCE OF THE IEEE ENGINEERING IN MEDICINE & BIOLOGY SOCIETY (EMBC)
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Ricky Maulana Fajri, Samaneh Khoshrou, Robert Peharz, Mykola Pechenizkiy
Summary: As social media plays a fixed role in our daily life, the issue of hostile contents and hate-speech is exacerbated, necessitating automatic hate-speech detection. A novel partition-based batch mode active learning framework is proposed to address the challenges of high class-skew, demonstrating substantial improvements in detection performance through extensive experiments.
MACHINE LEARNING AND KNOWLEDGE DISCOVERY IN DATABASES: APPLIED DATA SCIENCE AND DEMO TRACK, ECML PKDD 2020, PT V
(2021)
Proceedings Paper
Engineering, Electrical & Electronic
Alexander Fuchs, Johanna Rock, Mate Toth, Paul Meissner, Franz Pernkopf
Summary: Autonomous driving relies heavily on sensors to perceive the environment and deliver information to control systems. Radar sensors play a vital role in providing high resolution range and velocity measurements. However, the increased use of radar sensors in road traffic leads to mutual interference, which can be mitigated using Complex-Valued Convolutional Neural Networks (CVCNNs) to improve data efficiency and phase information conservation.
2021 IEEE RADAR CONFERENCE (RADARCONF21): RADAR ON THE MOVE
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
David Peter, Wolfgang Roth, Franz Pernkopf
Summary: This paper introduces the use of neural architecture search (NAS) to automatically discover small models for keyword spotting (KWS) in limited resource environments. By employing a differentiable NAS approach, the study achieves a highly efficient model with 95.4% accuracy on the Google speech commands dataset, with memory usage of 494.8 kB and 19.6 million operations. Additionally, weight quantization is utilized to reduce memory consumption further and improve model performance.
2020 25TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION (ICPR)
(2021)
Article
Computer Science, Artificial Intelligence
Rui Lv, Dingheng Wang, Jiangbin Zheng, Zhao-Xu Yang
Summary: In this paper, the authors investigate tensor decomposition for neural network compression. They analyze the convergence and precision of tensor mapping theory, validate the rationality of tensor mapping and its superiority over traditional tensor approximation based on the Lottery Ticket Hypothesis. They propose an efficient method called 3D-KCPNet to compress 3D convolutional neural networks using the Kronecker canonical polyadic (KCP) tensor decomposition. Experimental results show that 3D-KCPNet achieves higher accuracy compared to the original baseline model and the corresponding tensor approximation model.
Article
Computer Science, Artificial Intelligence
Xiangkun He, Zhongxu Hu, Haohan Yang, Chen Lv
Summary: In this paper, a novel constrained multi-objective reinforcement learning algorithm is proposed for personalized end-to-end robotic control with continuous actions. The approach trains a single model using constraint design and a comprehensive index to achieve optimal policies based on user-specified preferences.
Article
Computer Science, Artificial Intelligence
Zhijian Zhuo, Bilian Chen, Shenbao Yu, Langcai Cao
Summary: In this paper, a novel method called Expansion with Contraction Method for Overlapping Community Detection (ECOCD) is proposed, which utilizes non-negative matrix factorization to obtain disjoint communities and applies expansion and contraction processes to adjust the degree of overlap. ECOCD is applicable to various networks with different properties and achieves high-quality overlapping community detection.
Article
Computer Science, Artificial Intelligence
Yizhe Zhu, Chunhui Zhang, Jialin Gao, Xin Sun, Zihan Rui, Xi Zhou
Summary: In this work, the authors propose a Contrastive Spatio-Temporal Distilling (CSTD) approach to improve the detection of high-compressed deepfake videos. The approach leverages spatial-frequency cues and temporal-contrastive alignment to fully exploit spatiotemporal inconsistency information.
Review
Computer Science, Artificial Intelligence
Laijin Meng, Xinghao Jiang, Tanfeng Sun
Summary: This paper provides a review of coverless steganographic algorithms, including the development process, known contributions, and general issues in image and video algorithms. It also discusses the security of coverless steganography from theoretical analysis to actual investigation for the first time.
Article
Computer Science, Artificial Intelligence
Yajie Bao, Tianwei Xing, Xun Chen
Summary: Visual question answering requires processing multi-modal information and effective reasoning. Neural-symbolic learning is a promising method, but current approaches lack uncertainty handling and can only provide a single answer. To address this, we propose a confidence based neural-symbolic approach that evaluates NN inferences and conducts reasoning based on confidence.
Article
Computer Science, Artificial Intelligence
Anh H. Vo, Bao T. Nguyen
Summary: Interior style classification is an interesting problem with potential applications in both commercial and academic domains. This project proposes a method named ISC-DeIT, which combines data-efficient image transformer architectures and knowledge distillation, to address the interior style classification problem. Experimental results demonstrate a significant improvement in predictive accuracy compared to other state-of-the-art methods.
Article
Computer Science, Artificial Intelligence
Shashank Kotyan, Danilo Vasconcellos Vargas
Summary: This article introduces a novel augmentation technique called Dynamic Scanning Augmentation to improve the accuracy and robustness of Vision Transformer (ViT). The technique leverages dynamic input sequences to adaptively focus on different patches, resulting in significant changes in ViT's attention mechanism. Experimental results demonstrate that Dynamic Scanning Augmentation outperforms ViT in terms of both robustness to adversarial attacks and accuracy against natural images.
Article
Computer Science, Artificial Intelligence
Hiba Alqasir, Damien Muselet, Christophe Ducottet
Summary: The article proposes a solution to improve the learning process of a classification network by providing shape priors, reducing the need for annotated data. The solution is tested on cross-domain digit classification tasks and a video surveillance application.
Article
Computer Science, Artificial Intelligence
Dexiu Ma, Mei Liu, Mingsheng Shang
Summary: This paper proposes a method using neural dynamics solvers to solve infinity-norm optimization problems. Two improved solvers are constructed and their effectiveness and superiority are demonstrated through theoretical analysis and simulation experiments.
Article
Computer Science, Artificial Intelligence
Francesco Gregoretti, Giovanni Pezzulo, Domenico Maisto
Summary: Active Inference is a computational framework that uses probabilistic inference and variational free energy minimization to describe perception, planning, and action. cpp-AIF is a header-only C++ library that provides a powerful tool for implementing Active Inference for Partially Observable Markov Decision Processes through multi-core computing. It is cross-platform and improves performance, memory management, and usability compared to existing software.
Article
Computer Science, Artificial Intelligence
Zelin Ying, Dawei Cheng, Cen Chen, Xiang Li, Peng Zhu, Yifeng Luo, Yuqi Liang
Summary: This paper proposes a novel stock market trends prediction framework called SMART, which includes a self-supervised stock technical data sequence embedding model S3E. By training with multiple self-supervised auxiliary tasks, the model encodes stock technical data sequences into embeddings and uses the learned sequence embeddings for predicting stock market trends. Extensive experiments on China A-Shares market and NASDAQ market prove the high effectiveness of our model in stock market trends prediction, and its effectiveness is further validated in real-world applications in a leading financial service provider in China.
Article
Computer Science, Artificial Intelligence
Hao Li, Hao Jiang, Dongsheng Ye, Qiang Wang, Liang Du, Yuanyuan Zeng, Liu Yuan, Yingxue Wang, C. Chen
Summary: DHGAT1, a dynamic hyperbolic graph attention network, utilizes hyperbolic metric properties to embed dynamic graphs. It employs a spatiotemporal self-attention mechanism and weighted node representations, resulting in excellent performance in link prediction tasks.
Article
Computer Science, Artificial Intelligence
Jiehui Huang, Zhenchao Tang, Xuedong He, Jun Zhou, Defeng Zhou, Calvin Yu-Chian Chen
Summary: This study proposes a progressive learning multi-scale feature blending model for image deraining tasks. The model utilizes detail dilation and texture extraction to improve the restoration of rainy images. Experimental results show that the model achieves near state-of-the-art performance in rain removal tasks and exhibits better rain removal realism.
Article
Computer Science, Artificial Intelligence
Lizhi Liu, Zilin Gao, Yinhe Wang, Yongfu Li
Summary: This paper proposes a novel discrete-time interconnected model for depicting complex dynamical networks. The model consists of nodes and edges subsystems, which consider the dynamic characteristic of both nodes and edges. By designing control strategies and coupling modes, the stabilization and synchronization of the network are achieved. Simulation results demonstrate the effectiveness of the proposed methods.