Article
Computer Science, Artificial Intelligence
Chong Peng, Yiqun Zhang, Yongyong Chen, Zhao Kang, Chenglizhao Chen, Qiang Cheng
Summary: This paper proposes a new NMF method that enhances sparseness and robustness by imposing log-norm on the factor matrices and using a column-wisely sparse norm. Experimental results demonstrate the effectiveness of the proposed method.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Xiaoxia Zhang, Lu Chen, Ye Wang, Guoyin Wang
Summary: In this paper, a new recommendation algorithm called Three-way Decision Recommendations Based on Incremental Non-negative Matrix Factorization (3WD-INMF) is proposed, which leverages the concept of positive, negative, and boundary regions to update new samples' features. Experimental results show that the error induced by 3WD-INMF decreases with the addition of new samples and outperforms existing recommendation algorithms, indicating its superior performance and efficiency.
COGNITIVE COMPUTATION
(2022)
Article
Computer Science, Artificial Intelligence
Xiaoxia Zhang, Xianjun Zhou, Lu Chen, Yanjun Liu
Summary: Explicable recommendation systems are important for improving the persuasiveness of the system and enhancing user trust. However, the presence of latent features makes it challenging to interpret recommendation results. To address this, a novel method called PE-NMF is proposed, which replaces latent variables with explicit data to help users understand the features of recommended items and make better decisions. Experimental results demonstrate that PE-NMF performs well in rating prediction and top-N recommendation, outperforming FE-NMF and maintaining comparable recommendation ability to NMF.
ARTIFICIAL INTELLIGENCE REVIEW
(2023)
Article
Multidisciplinary Sciences
John Golden, Daniel O'Malley
Summary: This study found that combining forward annealing with reverse annealing in matrix factorization algorithms can significantly improve performance.
Article
Computer Science, Software Engineering
Yong Sheng Soh, Antonios Varvitsiotis
Summary: This paper introduces the application of the symmetric-cone multiplicative update algorithm to the cone factorization problem in the case of symmetric cones. The proposed algorithm updates each iterate by applying a chosen automorphism of the cone, ensuring that iterates remain within the interior of the cone. The algorithm utilizes a generalization of the geometric mean on symmetric cones. It has important applications in computing nonnegative matrix factorizations and hybrid lifts.
MATHEMATICAL PROGRAMMING
(2023)
Article
Environmental Sciences
Qin Jiang, Yifei Dong, Jiangtao Peng, Mei Yan, Yi Sun
Summary: The paper introduces a robust MLE-based NMF model for hyperspectral unmixing, which shows superior performance compared to existing NMF methods in experiments using simulated and real hyperspectral data sets.
Article
Computer Science, Information Systems
Xiaoxia Zhang, Degang Chen, Hong Yu, Guoyin Wang, Houjun Tang, Kesheng Wu
Summary: Nonnegative Matrix Factorization (NMF) produces interpretable solutions for applications like collaborative filtering. Regularization is needed to address issues like overfitting and interpretability. Existing regularizers are constructed from factorization results, but this study proposes a more holistic graph regularizer based on a linear projection of the rating matrix, named LPGNMF. Experimental results show the superiority of LPGNMF on different datasets.
INFORMATION SCIENCES
(2022)
Article
Environmental Sciences
Zhongliang Ren, Qiuping Zhai, Lin Sun
Summary: The emergence of hyperspectral imagery has opened up a new way for rapid mineral mapping. However, the classical spectral matching (SM) method suffers from poor mapping accuracy and inefficiency. To address these issues, we propose an unsupervised clustering-matching mapping method, using a combination of k-means and SM (KSM). Our study shows that KSM outperforms SM in terms of mapping accuracy and efficiency.
Article
Engineering, Multidisciplinary
Jiong Li, Lu Feng
Summary: This paper proposes a new signal processing framework composed of two stages: signal transformation and BSS algorithm for matrix estimation. Experimental results demonstrate the efficiency of the algorithm in estimating the mixing matrix.
MATHEMATICAL PROBLEMS IN ENGINEERING
(2021)
Article
Computer Science, Artificial Intelligence
Junjun Pan, Nicolas Gillis
Summary: Nonnegative matrix factorization (NMF) is a linear dimensionality technique for nonnegative data, which can be efficiently computed under the separability assumption. The algorithm operates by finding data points that contain basis vectors for decomposition.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2021)
Article
Engineering, Electrical & Electronic
Ying Zhang, Xiangli Li, Mengxue Jia
Summary: This paper introduces AGDNMF, an adaptive graph-based discriminative NMF method that utilizes label information to improve data representation and obtain the neighbor graph through adaptive iterations, which has been proven effective in various image datasets compared to state-of-the-art methods.
SIGNAL PROCESSING-IMAGE COMMUNICATION
(2021)
Article
Computer Science, Artificial Intelligence
Yuan Yuan, Zihan Zhang, Ganchao Liu
Summary: Hyperspectral remote sensing is an important method for earth observation, but the low spatial resolution of images makes it hard to distinguish ground objects. The HP-MLNMF framework proposed in this paper shows promising results in improving unmixing speed and efficiency in spectral signature estimation.
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
Chemistry, Multidisciplinary
Sebal Oo, Makoto Tsukai
Summary: This study aims to analyze the transition of foreign residents' characteristics with time series using a statistical approach to clarify policy building for foreign migrants by local government. A nonnegative matrix factorization model (NMF) was applied to cohort data of foreign residents in 47 Japanese prefectures in 2010, 2015, and 2020. The results showed significant differences in the cohort transition for foreign migrants with infants/children and the elderly, particularly between the Tohoku and Kyushu regions from 2010 to 2020 and the Tohoku region from 2015 to 2020. Support from the national government is necessary for the local governments in depopulated areas highlighted in this analysis due to their limited policy-building capacity.
APPLIED SCIENCES-BASEL
(2023)
Article
Computer Science, Artificial Intelligence
Guosheng Cui, Ye Li
Summary: This paper proposes a method called NRRNMF-ML, which reduces redundant information among multiple views and utilizes the manifold structure information of the data by introducing nonredundancy regularization and manifold regularization. Experimental results demonstrate the effectiveness of the proposed method.
INFORMATION FUSION
(2022)
Article
Biochemistry & Molecular Biology
Ernesto Aparicio-Puerta, Cristina Gomez-Martin, Stavros Giannoukakos, Jose Maria Medina, Chantal Scheepbouwer, Adrian Garcia-Moreno, Pedro Carmona-Saez, Bastian Fromm, Michiel Pegtel, Andreas Keller, Juan Antonio Marchal, Michael Hackenberg
Summary: The NCBI Sequence Read Archive hosts microRNA sequencing data for over 800 species, indicating a wide taxonomic distribution in small RNA research. With sRNAtoolbox, users can analyze an unlimited number of samples using high-confidence databases, enabling micro- and small RNA profiling.
NUCLEIC ACIDS RESEARCH
(2022)
Article
Biochemistry & Molecular Biology
Adrian Garcia-Moreno, Raul Lopez-Dominguez, Juan Antonio Villatoro-Garcia, Alberto Ramirez-Mena, Ernesto Aparicio-Puerta, Michael Hackenberg, Alberto Pascual-Montano, Pedro Carmona-Saez
Summary: This article presents a set of enrichment analysis methods for regulatory elements, including CpG sites, miRNAs, and transcription factors. Statistical significance is determined using a power weighting function for target genes and tested using the Wallenius noncentral hypergeometric distribution model to avoid selection bias. These new methodologies have shown promise in analyzing miRNAs associated with arrhythmia.
Article
Biochemical Research Methods
Daniel Toro-Dominguez, Jordi Martorell-Marugan, Manuel Martinez-Bueno, Raul Lopez-Dominguez, Elena Carnero-Montoro, Guillermo Barturen, Daniel Goldman, Michelle Petri, Pedro Carmona-Saez, Marta E. Alarcon-Riquelme
Summary: This study developed MyPROSLE, an omic-based analytical workflow, to measure the molecular characteristics of individual patients and support therapeutic decisions. Through analysis of nearly 6100 lupus patients and 750 healthy samples, it was found that dysregulation of certain gene-modules is strongly associated with specific clinical manifestations, relapses, long-term remission, and drug response. Therefore, MyPROSLE can accurately predict these clinical outcomes.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Computer Science, Theory & Methods
German Castano, Youssef Faqir-Rhazoui, Carlos Garcia, Manuel Prieto-Matias
Summary: This paper analyzes the Intel DPC++ Compatibility Tool and its performance, and studies the migration process of the Rodinia benchmarks, resulting in a high success rate of migration and moderate performance overhead.
JOURNAL OF PARALLEL AND DISTRIBUTED COMPUTING
(2022)
Article
Biology
Jordi Martorell-Marugan, Marco Chierici, Giuseppe Jurman, Marta E. Alarcon-Riquelme, Pedro Carmona-Saez
Summary: In this article, the potential of machine learning in the differential diagnosis of systemic lupus erythematosus and primary Sjogren's syndrome using gene expression and methylation data from 651 individuals is demonstrated. The impact of disease heterogeneity on predictive model performance is analyzed, with patients assigned to specific molecular clusters being misclassified more frequently and affecting overall model performance. Additionally, biomarkers that improve predictions compared to using the entire dataset are identified and validated with external studies.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Correction
Multidisciplinary Sciences
Matthew A. Reyna, David Haan, Marta Paczkowska, Lieven P. C. Verbeke, Miguel Vazquez, Abdullah Kahraman, Sergio Pulido-Tamayo, Jonathan Barenboim, Lina Wadi, Priyanka Dhingra, Raunak Shrestha, Gad Getz, Michael S. Lawrence, Jakob Skou Pedersen, Mark A. Rubin, David A. Wheeler, Sren Brunak, Jose M. G. Izarzugaza, Ekta Khurana, Kathleen Marchal, Christian von Mering, S. Cenk Sahinalp, Alfonso Valencia, Jueri Reimand, Joshua M. Stuart, Benjamin J. Raphael
NATURE COMMUNICATIONS
(2022)
Article
Computer Science, Software Engineering
Carlos Bilbao, Juan Carlos Saez, Manuel Prieto-Matias
Summary: Asymmetric multicore processors (AMPs) combine high-performance big cores and power-efficient small ones, sharing the same instruction set architecture but with different microarchitectural features. This trend indicates that AMPs may become a solid and more energy efficient replacement for symmetric multicores in various application domains. However, accessing hardware facilities for scheduling on AMPs may be challenging due to slow adoption in operating systems or incompatible forms in production systems. To address this, the PMCSched framework is proposed to simplify the development and evaluation of custom scheduling policies for multicore systems without patching the kernel. Experimental case studies on asymmetry-aware scheduling for Intel Alder Lake processors demonstrate the potential of the framework.
CONCURRENCY AND COMPUTATION-PRACTICE & EXPERIENCE
(2023)
Article
Biochemical Research Methods
Jordi Martorell-Marugan, Marco Chierici, Sara Bandres-Ciga, Giuseppe Jurman, Pedro Carmona-Saez
Summary: This study aims to conduct a comprehensive systematic review of studies applying machine learning to Parkinson's disease data. It summarizes the main advances in using machine learning methodologies for Parkinson's disease research and highlights the increasing interest in this area within the research community.
CURRENT BIOINFORMATICS
(2023)
Article
Computer Science, Theory & Methods
Carlos Bilbao, Juan Carlos Saez, Manuel Prieto-Matias
Summary: This paper analyzes how to combine smart partitioning of the last-level cache (LLC) and distributing threads across groups of cores to improve system-wide fairness. Based on the analysis, it proposes a fair OS-level NUMA-aware resource manager that utilizes dynamic contention-aware thread-to-socket mappings and cache-partitioning. The resource manager was implemented in the Linux kernel and evaluated on a real dual-socket system, showing an average reduction of unfairness by over 17% compared to Linux and a state-of-the-art NUMA-aware resource manager.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
(2023)
Article
Biochemistry & Molecular Biology
Sophia Mueller-Dott, Eirini Tsirvouli, Miguel Vazquez, Ricardo O. Ramirez Flores, Pau Badia-i-Mompel, Robin Fallegger, Denes Tuerei, Astrid Laegreid, Julio Saez-Rodriguez
Summary: Gene regulation plays a crucial role in cellular processes related to human health and disease. This study presents and evaluates a collection of reliable and comprehensive TF regulons created using the CollecTRI meta-resource. These regulons accurately estimate TF activities and help interpret transcriptomics data.
NUCLEIC ACIDS RESEARCH
(2023)
Proceedings Paper
Engineering, Electrical & Electronic
David Mallasen, Raul Murillo, Alberto A. Del Barrio, Guillermo Botella, Luis Pinuel, Manuel Prieto-Matias
Summary: The posit representation, or Unum-v3, is an alternative to the IEEE 754 standard that addresses the issues with floating-point numbers. The open-source PERCIVAL core integrates posit arithmetic and quire capabilities into hardware, while Xposit allows for compiling C programs with posit and quire instructions. PERCIVAL also supports IEEE 754 format for comparison purposes. This paper details the microarchitecture of the Posit Arithmetic Unit and describes the modifications made to the CVA6 core to add support for the Xposit RISC-V extension. FPGA synthesis results show the resource cost for supporting both posit with quire and IEEE 754 formats.
PROCEEDINGS OF THE 37TH CONFERENCE ON DESIGN OF CIRCUITS AND INTEGRATED SYSTEMS (DCIS 2022)
(2022)
Meeting Abstract
Rheumatology
Daniel Toro-Dominguez, Manuel Martinez-Bueno, Raul Lopez-Dominguez, Jordi Martorell-Marugan, Elena Carnero-Montoro, Guillermo Barturen, Daniel W. Goldman, Michelle Petri, Pedro Carmona-Saez, Marta Alarcon-Riquelme
ARTHRITIS & RHEUMATOLOGY
(2022)
Proceedings Paper
Computer Science, Interdisciplinary Applications
Manuel Costanzo, Enzo Rucci, Carlos Garcia-Sanchez, Marcelo Naiouf, Manuel Prieto-Matias
Summary: This paper presents the migration of a well-known sequence alignment tool, SW#db, from CUDA to DPC++ using the compatibility tool dpct provided by oneAPI. The experimental results demonstrate the usefulness of dpct for code migration and the cross-platform portability of the migrated DPC++ code, with similar or even better efficiency than the CUDA-native counterpart in some tests (approximately 5% improvement).
BIOINFORMATICS AND BIOMEDICAL ENGINEERING, PT II
(2022)
Article
Computer Science, Information Systems
David Mallasen, Raul Murillo, Alberto A. Del Barrio, Guillermo Botella, Luis Pinuel, Manuel Prieto-Matias
Summary: In this work, a PERCIVAL, an application-level posit RISC-V core based on CVA6, is presented, which can execute the complete posit instruction set, including quire fused operations. FPGA and ASIC synthesis reveal the high hardware cost of implementing 32-bit posits, but the quire enables more accurate execution of dot products. In matrix multiplications, using posits reduces accuracy error and maintains good execution speed.
IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTING
(2022)
