Article
Computer Science, Artificial Intelligence
Hai-Long Su, Zhi-Peng Li, Xiao-Bo Zhu, Li-Na Yang, Valeriya Gribova, Vladimir Fedorovich Filaretov, Anthony G. Cohn, De-Shuang Huang
Summary: In this article, we propose a novel GNN model based on semi-implicit VI (SIVI) to handle relational data with uncertain properties. The model improves VI flexibility and expressiveness by embedding nodes into a latent space and using a mixing distribution. Experimental results show that our method achieves state-of-the-art results compared to other similar methods on multiple data sets.
IEEE TRANSACTIONS ON COGNITIVE AND DEVELOPMENTAL SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Haoliang Sun, Xiankai Lu, Haochen Wang, Yilong Yin, Xiantong Zhen, Cees G. M. Snoek, Ling Shao
Summary: This paper proposes attentional prototype inference (API), a probabilistic latent variable framework for few-shot segmentation, to address the uncertainty and ambiguity caused by limited labeled examples. By modeling the prototype as a probabilistic distribution, the model's generalization ability is enhanced. In addition, the introduction of a local latent variable representing the attention map further improves the model's performance. Experimental results show that the proposed method achieves competitive or better performance compared to state-of-the-art prototype-based methods.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Information Systems
Hatem Khalloof, Mohammad Mohammad, Shadi Shahoud, Clemens Duepmeier, Veit Hagenmeyer
Summary: Parallelizing population-based metaheuristics, such as Evolutionary Algorithms, on a cluster computing environment can significantly enhance computational performance. This article presents a scalable software architecture that allows users to combine basic parallelization models of EAs efficiently, showing high potential for scaling up optimization speed of complex problems. Extensive evaluation combining Coarse-Grained and Global Models for solving optimization problems demonstrates the effectiveness of the approach.
INTERNET OF THINGS
(2021)
Article
Computer Science, Artificial Intelligence
Divyang Teotia, Agata Lapedriza, Sarah Ostadabbas
Summary: This paper presents Hierarchical Network Dissection, a general pipeline for interpreting the internal representation of face-centric inference models. It addresses the challenges of overlapping concepts and global concepts in face-centric models, and reveals biases and characteristics of both the models and the training data through interpretability results.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2022)
Article
Engineering, Mechanical
Alejandro Poblete, Rafael O. Ruiz, Gaofeng Jia
Summary: This work proposes a hierarchical Bayesian framework to identify the electromechanical properties and uncertainties of Piezoelectric Energy Harvesters (PEHs) based on experimental frequency response functions (FRFs). The framework explicitly models the dispersion in FRF observed in groups of PEHs as a consequence of uncertainties in the model parameters. The proposed framework extracts more information about the model parameters and accounts for the uncertainties across different devices.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2022)
Article
Computer Science, Artificial Intelligence
Wei Li, Shaogang Gong, Xiatian Zhu
Summary: This study addresses the scalability problem of person search by proposing a Hierarchical Distillation Learning (HDL) approach that distils knowledge from a strong teacher model to a lightweight student model. The design of a powerful teacher model for joint learning in unconstrained scene images demonstrates the modeling advantages and cost-effectiveness superiority of HDL over existing state-of-the-art person search methods.
PATTERN RECOGNITION
(2021)
Article
Environmental Sciences
Shutaro Shiraki, Aung Kyaw Thu, Yutaka Matsuno, Yoshiyuki Shinogi
Summary: The two-layer Shuttleworth-Wallace (SW) model has been widely used for predicting evapotranspiration (ET) with good results. By comparing different Bayesian approaches, it was found that a two-level hierarchical Bayesian (HB) approach showed better performance in parameter estimation compared to a simple non-hierarchical Bayesian (SB) approach. This highlights the importance of considering seasonal fluctuations and variation in crop growth stages for accurate parameter estimation.
Article
Mathematics, Applied
Thierno Souleymane Barry, Oscar Ngesa, Jeremiah Kimani Kiingati, Nelson Owuor Onyango, Aurise Niyoyunguruza, Alexis Habineza, Henry Mwambi, Henri Bello Fika
Summary: Anemia and malaria are the leading causes of morbidity and mortality among children under five years old in sub-Saharan Africa. This study aims to estimate the spatial linear correlation between anemia and malaria, and investigate the factors affecting morbidity in Guinea. The findings show high prevalences of anemia and malaria in children under five years old in Guinea, and significant associations between each disease and various demographic factors.
Article
Computer Science, Artificial Intelligence
Ivona Najdenkoska, Xiantong Zhen, Marcel Worring, Ling Shao
Summary: Automating report generation for medical imaging using probabilistic variational topic inference can generate reports with novel sentence structure, rather than mere copies of training samples, while achieving comparable performance to state-of-the-art methods.
MEDICAL IMAGE ANALYSIS
(2022)
Article
Green & Sustainable Science & Technology
Lixuan Chen, Tianyu Mu, Xiuting Li, Jichang Dong
Summary: In recent years, China's population growth rate has been declining. Accurately predicting population development trends has become a top priority to facilitate national planning and decision making. Using multiple models, this study finds that the population gap between cities is widening. Geographically, the population growth rate is balanced between northern and southern China. High-tier cities are experiencing slower population growth, low-tier cities have a negative growth, while middle-tier cities are seeing a skyrocketing population growth. Despite regional integration, the radiative driving effect of large urban agglomerations and metropolitan areas is relatively limited.
Article
Public, Environmental & Occupational Health
Tsigereda Tilahun Letta, Denekew Bitew Belay, Endale Alemayehu Ali
Summary: This study used a Bayesian hierarchical model to identify risk factors associated with cholera disease in Ethiopia. The results showed that admission status, age, presence of another sick person in the family, dehydration status, oral rehydration salt, intravenous, and antibiotics were significantly associated with cholera outbreak disease.
Article
Computer Science, Information Systems
Lorena Romero-Medrano, Pablo Moreno-Munoz, Antonio Artes-Rodriguez
Summary: Bayesian change-point detection with latent variable models improves detection accuracy and reduces delay in high-dimensional time-series. The proposed multinomial sampling method allows for stable and analytically tractable inference, outperforming baseline methods. The results show promise for various applications, including human behavioral studies.
JOURNAL OF SIGNAL PROCESSING SYSTEMS FOR SIGNAL IMAGE AND VIDEO TECHNOLOGY
(2022)
Article
Neurosciences
Seyedeh-Rezvan Farahibozorg, Janine D. Bijsterbosch, Weikang Gong, Saad Jbabdi, Stephen M. Smith, Samuel J. Harrison, Mark W. Woolrich
Summary: The study introduces a new technique named sPROFUMO for estimating functional brain networks at the individual level from large-scale brain imaging datasets, with high predictive potential. Through simulations and empirical analysis on UKB subjects, it demonstrates the potential and advantages of this model in investigating individualized brain function profiles.
Article
Engineering, Mechanical
Junming Ma, Nani Bai, Yi Zhou, Chengming Lan, Hui Li, B. F. Spencer
Summary: This article proposes the use of generalized hierarchical Bayesian inference for fatigue life prediction based on general multi-parameter Weibull models. The article establishes a three-layer hierarchical Bayesian structure and uses Gibbs sampling to obtain posterior samples for parameters and hyperparameters. The results show that the scatter in fatigue life prediction for the corroded specimens becomes smaller when considering informative priors for the parameters in the Weibull model.
INTERNATIONAL JOURNAL OF FATIGUE
(2022)
Article
Engineering, Mechanical
Xinyu Jia, Wang-Ji Yan, Costas Papadimitriou, Ka-Veng Yuen
Summary: The hierarchical Bayesian modelling (HBM) framework is a recent method proposed to account for model parameter uncertainty in structural dynamics. This study employs a variational inference scheme to derive explicit expressions for the posterior distributions of the hyper parameters and the predictive distribution of the model parameters, enhancing the computational efficiency of the HBM framework.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)