Article
Biology
Natascha Ingrid Drude, Lorena Martinez Gamboa, Meggie Danziger, Ulrich Dirnagl, Ulf Toelch
Summary: The purpose of preclinical research is to guide the development of novel diagnostics or therapeutics, with results from animal models informing decisions on human studies. Despite apparent efficacy demonstrated in preclinical studies, a significant number of clinical trials still fail. Large-scale replication studies are currently investigating factors influencing the robustness of preclinical research.
Article
Multidisciplinary Sciences
Aurelien Allard, Christine Clavien
Summary: The recent replicability crisis in social and biomedical sciences has emphasized the importance of improving the honest transmission of scientific content. Two studies were conducted to investigate whether nudges and soft social incentives could enhance participants' readiness to transmit high-quality scientific news. The findings indicate that although participants have a preference for studies using high sample sizes and randomized designs, they are biased towards positive results and prefer results that align with their previous intuitions (confirmation bias).
Article
Psychology, Biological
Alex D. McDiarmid, Alexa M. Tullett, Cassie M. Whitt, Simine Vazire, Paul E. Smaldino, Jeremy E. Stephens
Summary: The study found that psychologists did update their beliefs about effect sizes after learning about new evidence from replication studies, although not as much as predicted by a rational Bayesian model. They also seemed to underestimate the evidentiary value of replication studies.
NATURE HUMAN BEHAVIOUR
(2021)
Article
Computer Science, Information Systems
Soufiane Khedairia, Mohamed Tarek Khadir
Summary: This paper introduces an Iterative Combining Clusterings Method (ICCM) that iteratively processes the dataset by extracting sub-clusters through a voting process, achieving higher effectiveness and robustness. Experimental results show significant improvements in clustering quality metrics and external validation metrics, confirming the usefulness of the proposed approach compared to other clustering ensemble methods.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Xinwang Liu
Summary: The newly proposed localized simple multiple kernel k-means (SimpleMKKM) provides an elegant clustering framework that considers the potential variation among samples. However, it requires pre-specifying an extra hyperparameter, limiting its practical applications. To overcome this issue, a hyperparameter-free localized SimpleMKKM is proposed, which jointly learns the optimal coefficient of neighborhood mask matrices together with the clustering tasks. The obtained optimum is proved to be the global one, and comprehensive experimental studies verify its effectiveness.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Multidisciplinary Sciences
Hannah M. Fraser, Martin Bush, Bonnie Wintle, Fallon G. Mody, Eden Smith, Anca P. Hanea, Elliot Gould, Victoria Hemming, Daniel Hamilton, Libby Rumpff, David T. Wilkinson, Ross Pearson, Felix Singleton Thorn, Raquel Ashton, Aaron Willcox, Charles Gray, Andrew H. Head, Melissa Ross, Rebecca Groenewegen, Alexandru R. Marcoci, Ans Vercammen, Timothy Parker, Rink O. Hoekstra, Shinichi Nakagawa, David Mandel, Don van Ravenzwaaij, Marissa McBride, Richard Sinnott, Peter Vesk, Mark Burgman, Fiona Fidler
Summary: In order to address the resource-intensive nature of replicating individual studies, techniques for predicting replicability are necessary. The repliCATS process is introduced as a structured expert elicitation approach, based on modified Delphi technique, to evaluate the replicability of research claims in social and behavioral sciences. This process has shown high accuracy in prediction, scalability, and potential to collect qualitative data for understanding the limits of generalizability. The main limitation is its reliance on human-derived predictions, which may lead to participant fatigue.
Article
Computer Science, Information Systems
Mengjing Sun, Siwei Wang, Pei Zhang, Xinwang Liu, Xifeng Guo, Sihang Zhou, En Zhu
Summary: Multiple kernel subspace clustering (MKSC) is an important extension for handling multi-view non-linear subspace data. It aims to build a flexible and appropriate graph for clustering from the kernel space. However, existing MKSC methods suffer from noise and unreliable similarity measures in the original high-dimensional spaces, resulting in low-quality graph matrices and degraded clustering performance. Inspired by projective clustering, this paper proposes a projective multiple kernel subspace clustering (PMKSC) method, which fuses multiple kernel graphs in the low-dimensional partition space to alleviate noise and redundancy and obtain high-quality similarity for uncovering underlying clustering structures.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)
Article
Computer Science, Artificial Intelligence
Mohammad Sultan Mahmud, Joshua Zhexue Huang, Salvador Garcia
Summary: This study proposes a new distributed clustering approximation framework for big data, which uses multiple random samples to compute an ensemble result and integrates component clustering results using two new methods. Experimental results demonstrate the accuracy in identifying the correct number of clusters and the better scalability, efficiency, and clustering stability of the proposed methods.
INFORMATION FUSION
(2024)
Article
Health Care Sciences & Services
Tiffany Dal Santo, Danielle B. Rice, Lara S. N. Amiri, Amina Tasleem, Kexin Li, Jill T. Boruff, Marie-Claude Geoffroy, Andrea Benedetti, Brett D. Thombs
Summary: We investigated recent meta-research studies on adherence to reporting guidelines and found that most studies reported suboptimal adherence in health research. However, a small proportion of these studies provided enough information for verification or replication.
JOURNAL OF CLINICAL EPIDEMIOLOGY
(2023)
Article
Biology
Ran Dai, Cheng Zheng
Summary: False discovery rate (FDR) controlling procedures ensure statistical guarantees for replicability in signal identification based on multiple hypotheses testing. This paper introduces a knockoff-based variable selection method (Simultaneous knockoffs) that provides exact FDR control guarantees under finite sample settings. This method identifies mutual signals from multiple independent datasets by jointly considering information from different sources (with potential heterogeneity).
Article
Computer Science, Artificial Intelligence
Antonio Cavalcante Araujo Neto, Jorg Sander, Ricardo J. G. B. Campello, Mario A. Nascimento
Summary: HDBSCAN*, a cutting-edge hierarchical clustering method based on density, can be challenging to optimize for the parameter mpts. This paper introduces an approach that allows for the efficient computation of multiple HDBSCAN* hierarchies for a range of mpts values, achieving over a hundred hierarchies for a computational cost equivalent to running HDBSCAN* twice.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2021)
Article
Computer Science, Artificial Intelligence
Nan Zhang, Shiliang Sun
Summary: Multiview clustering is an important research topic, and incomplete views of data instances are common in real-world scenarios. To address this issue, we propose an effective incomplete multiview nonnegative representation learning framework that can handle incomplete multiview clustering in various situations and achieves better results compared to other state-of-the-art algorithms.
PATTERN RECOGNITION
(2022)
Article
Computer Science, Artificial Intelligence
Mario Garza-Fabre, Julia Handl, Adan Jose-Garcia
Summary: Multiview data analysis integrates distinct information sources in many applications. Data clustering in a multiview setting aims to group entities based on multiple perspectives. This article proposes a new evolutionary method for multiview clustering, optimizing multiple objectives simultaneously. Experimental evaluation demonstrates the effectiveness of the proposed method in discovering high-quality partitions, considering a bioinformatics application and various synthetic problems. The method also exhibits robustness against unreliable data sources and automatic determination of the number of clusters.
IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION
(2023)
Article
Computer Science, Artificial Intelligence
Xinwang Liu
Summary: In this paper, we propose a simple yet effective multiple kernel clustering algorithm called SimpleMKKM. The algorithm extends the supervised kernel alignment criterion to multi-kernel clustering and solves an intractable minimization-maximization problem to obtain the clustering results. The experimental study demonstrates that SimpleMKKM outperforms state-of-the-art multiple kernel clustering alternatives in terms of clustering accuracy, formulation and optimization advantages, and other aspects.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Automation & Control Systems
Lan Bai, Yuan-Hai Shao, Zhen Wang, Wei-Jie Chen, Nai-Yang Deng
Summary: Cross-manifold clustering is a challenging learning problem, and traditional clustering methods often fail in this scenario. The proposed multiple flat projections clustering (MFPC) effectively discovers global structures of implicit manifolds by projecting samples into localized flats and solving nonconvex matrix optimization problems through a recursive algorithm. The nonlinear version of MFPC, extended via kernel tricks, shows promising results in dealing with complex cross-manifold learning situations.
IEEE TRANSACTIONS ON CYBERNETICS
(2022)