Article
Biochemical Research Methods
Antoine Passemiers, Yves Moreau, Daniele Raimondi
Summary: This article presents a novel method called PORTIA for inferring gene regulatory networks (GRNs). The method is based on robust precision matrix estimation and is shown to outperform state-of-the-art methods in terms of speed while still maintaining good accuracy. The authors extensively validated PORTIA using benchmark datasets and propose a new scoring metric based on graph-theoretical concepts.
Article
Computer Science, Information Systems
Tingrui Liu, Xin Li, Liguo Tan, Shenmin Song
Summary: This study proposes an algorithm based on an incremental learning model for multiobjective estimation of distributions. The algorithm incorporates an adaptive learning mechanism to discover the structure of the Pareto-optimal set during evolutionary search. Experimental results demonstrate a significant improvement over several benchmark tests.
INFORMATION SCIENCES
(2021)
Article
Computer Science, Artificial Intelligence
Zhen Peng, Minnan Luo, Wenbing Huang, Jundong Li, Qinghua Zheng, Fuchun Sun, Junzhou Huang
Summary: This paper investigates how to extract abundant information from graph-structured data into embedding space without external supervision. The authors propose a novel concept, Graphical Mutual Information (GMI), to measure the correlation between input graph and hidden representation. They develop an unsupervised embedding model based on GMI and apply it to the anomaly detection task, achieving promising performance in various downstream tasks.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Information Systems
Pei Wang, Yun Yang, Yuelong Xia, Kun Wang, Xingyi Zhang, Song Wang
Summary: This paper proposes an information maximization adaptation network with label distribution priors to address the challenges brought by pseudo labels in unsupervised domain adaptation. By maximizing source mutual information, introducing weighted target mutual information, and adding a regularization term of label priors distribution, this method achieves remarkable results on three benchmark datasets.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Computer Science, Software Engineering
Zhiyi Lin, Qing Su, Guobo Xie
Summary: The NMIEDA algorithm is proposed to overcome premature convergence in bivariate estimation algorithms by using normalized mutual information to measure variable interaction and generating a dependency forest model. It provides a new updating mechanism based on sporadic model building and a reward and punishment scheme, and adopts a new sampling mechanism to improve efficiency by combining stochastic sampling, opposition-based learning, and mutation operators. Simulation results show that NMIEDA outperforms other bivariate algorithms on benchmark and real-world problems.
CONCURRENCY AND COMPUTATION-PRACTICE & EXPERIENCE
(2021)
Article
Computer Science, Artificial Intelligence
Yangyi Du, Xiaojun Zhou, Chunhua Yang, Tingwen Huang
Summary: In this paper, an interactive feature selection framework based on state transition algorithm is proposed to address high-dimensional feature selection problems. The framework combines the advantages of filter and wrapper methods to improve classification efficiency, and employs self-adaptive mechanism and multi-step STA to avoid local optima.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Biochemical Research Methods
Haonan Feng, Ruiqing Zheng, Jianxin Wang, Fang-Xiang Wu, Min Li
Summary: Gene regulatory networks play a crucial role in biological processes, and existing expression data can be used to infer these networks using computational methods. However, identifying indirect regulatory links remains a challenge. In this study, a novel information-theory-based method is proposed to improve the identification of regulatory relationships between genes.
IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS
(2022)
Article
Physics, Multidisciplinary
Xiyu Shi, Varuna De-Silva, Yusuf Aslan, Erhan Ekmekcioglu, Ahmet Kondoz
Summary: This paper reveals the typical learning patterns of convolutional neural networks using information theoretical measures, showing that more convolutional layers improve learning but excessively adding layers does not. It also demonstrates that the kernel size of convolutional layers only affects learning speed, and the placement of dropout layers has varying effects depending on the dropout rate.
Article
Computer Science, Information Systems
Ozan Oezdenizci, Deniz Erdogmus
Summary: This study introduces an emerging information theoretic feature transformation protocol as an end-to-end neural network training approach, achieving feature dimensionality reduction and experimental evaluation on high-dimensional biological data sets.
INFORMATION SCIENCES
(2021)
Article
Biochemistry & Molecular Biology
Evan Maltz, Roy Wollman
Summary: Quantifying the dependency between mRNA abundance and downstream cellular phenotypes is a fundamental problem in biology. In this study, multimodal single-cell measurement data was used to analyze the expression of 83 genes in the Ca2+ signaling network and the dynamic Ca2+ response. It was found that the overall expression levels of these genes explain approximately 60% of Ca2+ signal entropy, with each single gene contributing an average of 17% and showing a large degree of redundancy. The study also estimated the dependency between the size of a gene set and its information content, revealing that on average, a set of 53 genes contains 54% of the information about Ca2+ signaling.
MOLECULAR SYSTEMS BIOLOGY
(2022)
Article
Thermodynamics
Min Li, Yi Yang, Zhaoshuang He, Xinbo Guo, Ruisheng Zhang, Bingqing Huang
Summary: This paper focuses on constructing accurate wind speed forecasting models and quantitatively analyzing the models using interpretable analysis. A multivariate wind speed forecasting model (PMI-CMOGSA-RELM) is proposed based on machine learning methods and a clustering-based multi-objective gravity search algorithm (CMOGSA). A new evaluation metric, Absolute Error Coverage Probability (AECP), is proposed to better evaluate forecasting accuracy. Post-hoc attribution analysis methods and visualization tools are used to analyze the interpretability of the forecasting model. The experimental results validate the proposed model and demonstrate its smaller errors, higher estimation accuracy, and better understandability.
Article
Engineering, Electrical & Electronic
Mauricio E. Gonzalez, Jorge F. Silva, Miguel Videla, Marcos E. Orchard
Summary: This work presents a method for testing the independence of two continuous and finite-dimensional random variables using a data-driven partition. By approximating the sufficient statistics of an oracle test, a learning criterion is provided for designing the partition. The method achieves a consistent and distribution-free test of independence over the family of probabilities with a density.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2022)
Article
Biology
Kelsey E. Huntington, Anna D. Louie, Chun Geun Lee, Jack A. Elias, Eric A. Ross, Wafik S. El-Deiry
Summary: A study found that a small number of clinical and cytokine expression variables are predictive of COVID-19 disease severity, questioning the mechanism by which the virus causes severe illness.
Article
Computer Science, Artificial Intelligence
Na Dong, Yongqiang Zhang, Mingli Ding, Yancheng Bai
Summary: Deep learning architectures perform well in object detection, but suffer from performance drop when learning new classes incrementally without forgetting old ones. Many incremental learning methods have been proposed to address the catastrophic forgetting problem, but current strategies encounter issues in preserving old knowledge and learning new classes simultaneously.
PATTERN RECOGNITION
(2023)
Article
Biochemical Research Methods
Lior I. I. Shachaf, Elijah Roberts, Patrick Cahan, Jie Xiao
Summary: In this study, a new method for gene regulatory network reconstruction is proposed, which combines CMIA and the KSG-MI estimator. The results show that this method achieves an improvement of 20-35% in precision-recall measures compared to the current gold standard. This new method will help researchers discover new gene interactions or better choose gene candidates for experimental validations.
BMC BIOINFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Feng Wang, Cheng Yang, Zhiyi Lin, Yuanxiang Li, Yuan Yuan
Proceedings Paper
Computer Science, Artificial Intelligence
Yifei Xu, Jingqiao Zhang, Ru He, Liangzhu Ge, Chao Yang, Cheng Yang, Ying Nian Wu
Summary: The core of self-supervised learning for pre-training language models lies in the design of pre-training tasks as well as appropriate data augmentation. This paper proposes a self-augmentation strategy (SAS) that utilizes a single network for both regular pre-training and contextualized data augmentation, outperforming ELECTRA and other state-of-the-art models in GLUE tasks with similar or less computation cost.
THIRTY-SIXTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE / THIRTY-FOURTH CONFERENCE ON INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE / TWELVETH SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE
(2022)
Proceedings Paper
Acoustics
Jihai Zhang, Fangquan Lin, Wei Jiang, Cheng Yang, Gaoge Liu
Summary: With the rapid development of e-commerce, it has become essential but challenging to provide a recommending service for users quickly. This paper proposes a novel embedding-based method called NATM, which incorporates both graph-based and sequential information to improve the retrieval stage of the recommender system, aiming to enhance the accuracy and effectiveness of recommendations.
2022 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP)
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Ziqiang Cui, Yixin Su, Fangquan Lin, Cheng Yang, Hanwei Zhang, Jihai Zhang
Summary: This paper proposes a Dual Disentangled Attention (DDA) based BERT model, called DDA-BERT, to better leverage multi-information in sequential recommendation systems. Extensive experiments on three benchmark datasets demonstrate that DDA-BERT consistently outperforms the state-of-the-art baselines by up to 30%.
2022 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN)
(2022)
Proceedings Paper
Computer Science, Information Systems
Wei Wang, Liangzhu Ge, Jingqiao Zhang, Cheng Yang
Summary: This study improves the performance of contrastive learning in unsupervised sentence embeddings by introducing switch-case augmentation and sampling hard negatives from a pre-trained language model, achieving significant results on STS benchmarks.
PROCEEDINGS OF THE 45TH INTERNATIONAL ACM SIGIR CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION RETRIEVAL (SIGIR '22)
(2022)
Proceedings Paper
Computer Science, Information Systems
Jihai Zhang, Fangquan Lin, Cheng Yang, Wei Wang
Summary: CTR prediction plays a crucial role in modeling recommender systems. Previous studies have mainly focused on user behavior modeling, neglecting the representations of candidate items. In this paper, we propose a Deep multi-Representational Item NetworK (DRINK) that addresses the sparse user behavior problem and captures the multi-representational characteristics of candidate items using a transformer-based approach. We also decouple time information and item behavior to avoid information overwhelming. Experimental results on real-world datasets demonstrate the effectiveness of the proposed model.
PROCEEDINGS OF THE 45TH INTERNATIONAL ACM SIGIR CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION RETRIEVAL (SIGIR '22)
(2022)
Proceedings Paper
Computer Science, Information Systems
Jihai Zhang, Fangquan Lin, Cheng Yang, Wei Jiang
Summary: This paper proposes a transformer-based spatial-temporal recommendation framework (STEM), which utilizes attention mechanisms and a transformer-based model to incorporate user behavior, item behavior, and spatial-temporal information for improving recommendation performance.
PROCEEDINGS OF THE 45TH INTERNATIONAL ACM SIGIR CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION RETRIEVAL (SIGIR '22)
(2022)
Proceedings Paper
Computer Science, Information Systems
Wei Jiang, Fangquan Lin, Jihai Zhang, Cheng Yang, Hanwei Zhang, Ziqiang Cui
Summary: This research introduces a novel time-aware framework for dynamic sequential recommendation, addressing the issue of dynamic property in user behavior modeling. The framework consists of a time-invariant main network and a time-sensitive bias network, enabling the capture of both sequential and temporal patterns simultaneously.
21ST IEEE INTERNATIONAL CONFERENCE ON DATA MINING WORKSHOPS ICDMW 2021
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Junhao Hua, Ling Yan, Huan Xu, Cheng Yang
Summary: This paper introduces a novel data-driven and interpretable pricing approach for markdowns, aiming to maximize the overall profit of perishable products over their finite selling horizon by leveraging counterfactual prediction and multi-period price optimization.
KDD '21: PROCEEDINGS OF THE 27TH ACM SIGKDD CONFERENCE ON KNOWLEDGE DISCOVERY & DATA MINING
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Mingdong Ou, Nan Li, Cheng Yang, Shenghuo Zhu, Rong Jin
THIRTY-THIRD AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE / THIRTY-FIRST INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE / NINTH AAAI SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE
(2019)
Proceedings Paper
Computer Science, Artificial Intelligence
Yu-Hang Zhou, Chen Liang, Nan Li, Cheng Yang, Shenghuo Zhu, Rong Jin
THIRTY-THIRD AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE / THIRTY-FIRST INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE / NINTH AAAI SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE
(2019)
Proceedings Paper
Computer Science, Artificial Intelligence
Kui Zhao, Yuechuan Li, Zhaoqian Shuai, Cheng Yang
KDD'18: PROCEEDINGS OF THE 24TH ACM SIGKDD INTERNATIONAL CONFERENCE ON KNOWLEDGE DISCOVERY & DATA MINING
(2018)