Article
Computer Science, Artificial Intelligence
Xiaohuan Lu, Jiang Long, Jie Wen, Lunke Fei, Bob Zhang, Yong Xu
Summary: In this paper, a novel method called LPP_SGE is proposed for unsupervised dimensionality reduction. LPP_SGE introduces a novel adaptive graph learning model and obtains the intrinsic graph and projection in a unified framework by fully exploring the representation information and distance information of the original data. It simultaneously captures the representation information and distance information in one term. Moreover, LPP_SGE enhances the robustness by introducing an 'l2,1' norm based projection constraint to select the most discriminative features from the complex data.
PATTERN RECOGNITION
(2022)
Article
Automation & Control Systems
Zhixiao Wang, Yahui Chai, Chengcheng Sun, Xiaobin Rui, Hao Mi, Xinyu Zhang, Philip S. Yu
Summary: This paper proposes a weighted symmetric graph embedding approach for link prediction, aiming to solve the problems of node embedding and edge embedding. In node embedding, different aggregating weights are used to aggregate neighbors in different orders. In edge embedding, bidirectional concatenation is employed to ensure the symmetry of edge representations while preserving local structural information. Experimental results demonstrate that the proposed approach outperforms state-of-the-art methods in predicting network links.
IEEE TRANSACTIONS ON CYBERNETICS
(2022)
Article
Computer Science, Artificial Intelligence
Qian Li, Daling Wang, Shi Feng, Cheng Niu, Yifei Zhang
Summary: This article introduces a novel global graph attention embedding network (GGAE) for relation prediction, which combines global information from both direct neighbors and multihop neighbors to obtain better embeddings of entities and relations. The model utilizes path construction algorithms, path modeling methods, and graph attention mechanisms to capture information from neighbors at different distances, enhancing the comprehensive semantic information of entities and relations.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Shichao Wu, Lei Zhou, Zhengxi Hu, Jingtai Liu
Summary: Research shows that the hierarchical context-based emotion recognition method using scene graphs can improve emotion recognition accuracy in unconstrained environments, resulting in significant performance enhancements.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Yuenan Li, Qixin Yan, Kuangshi Zhang, Haoyu Xu
Summary: In this paper, a deep neural network is proposed for single image reflection removal, which utilizes a pyramid structure and grid module design to improve performance. The network leverages multi-scale feature learning and task-driven regularization strategy to enhance its effectiveness.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2022)
Article
Mechanics
Giovanni Piccioli, Guilhem Semerjian, Gabriele Sicuro, Lenka Zdeborova
Summary: The study investigates a polynomial time message-passing algorithm designed to solve the inference problem of partially recovering the hidden permutation, in the sparse regime with constant average degrees. Extensive numerical simulations are conducted to determine the range of parameters in which this algorithm achieves partial recovery, and the algorithm is also extended to a generalized ensemble of correlated random graphs with prescribed degree distributions.
JOURNAL OF STATISTICAL MECHANICS-THEORY AND EXPERIMENT
(2022)
Article
Computer Science, Information Systems
Anran Li, Shuangshuang Xue, Xiang-Yang Li, Lan Zhang, Jianwei Qian
Summary: In this work, a framework called AppDNA is designed to automatically generate a compact representation for each app in order to comprehensively analyze its behaviors. The app representation can be used for various objectives such as malware detection, app categorization, and app version detection. The proposed approach utilizes a function-call-graph-based app profiling scheme to convert a large function call graph into a fixed-length vector for robust app profiling. Extensive evaluations demonstrate that the approach achieves high accuracy and low computation cost for different tasks.
IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTING
(2022)
Article
Mathematics, Applied
Chuanyuan Ge, Shiping Liu
Summary: This paper establishes nodal domain theorems for arbitrary symmetric matrices by studying the induced signed graph structure. Our definitions of nodal domains for any function on a signed graph are switching invariant. When the induced signed graph is balanced, our definitions and upper bound estimates reduce to existing results for generalized Laplacians.
CALCULUS OF VARIATIONS AND PARTIAL DIFFERENTIAL EQUATIONS
(2023)
Article
Mathematics
Gi-Sang Cheon, Jang Soo Kim, Seyed Ahmad Mojallal, Meesue Yoo
Summary: The paper examines the spectral properties of the symmetric Pascal matrix and binomial graph, including eigenvalues, eigenvectors, algebraic connectivity, and inertia indices. The determinant of the Pascal matrix modulo 3 is also computed in the study.
LINEAR & MULTILINEAR ALGEBRA
(2022)
Article
Engineering, Multidisciplinary
Konstantinos Tsitseklis, Maria Krommyda, Vasileios Karyotis, Verena Kantere, Symeon Papavassiliou
Summary: Community detection utilizing hyperbolic network embedding is demonstrated in this paper, expanded with a graph database approach for scalability. The method is applicable for analyzing large datasets and visualizing communities in RDF data and linked datasets from diverse areas, showcasing its feasibility in producing correct results while scaling seamlessly.
IEEE TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING
(2021)
Article
Mathematics, Applied
Rupert H. Levene, Polona Oblak, Helena Smigoc
Summary: This article introduces the notion of compatibility for multiplicity matrices and presents a necessary condition for the join of two graphs to be the pattern of an orthogonal symmetric matrix. The authors also prove that under certain additional hypotheses, this necessary condition is sufficient. As an application, they demonstrate that when the two graphs are unions of complete graphs, the minimum number of distinct eigenvalues is either two or three, and provide the characterisation for each case.
LINEAR ALGEBRA AND ITS APPLICATIONS
(2022)
Article
Engineering, Biomedical
Jitao Zhong, Wenyan Du, Lu Zhang, Hong Peng, Bin Hu
Summary: Automatic detection of depression is crucial in today's society. This paper proposes an automatic feature extraction method called Sparse Graphs Embedding (SGE) for depression detection. The method addresses key challenges and achieves promising detection rates using fNIRS.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2023)
Article
Engineering, Electrical & Electronic
Nir Shlezinger, Nariman Farsad, Yonina C. Eldar, Andrea J. Goldsmith
Summary: This paper proposes a framework that combines model-based algorithms and data-driven machine learning methods for inference from stationary time sequences. By learning specific components of the factor graph representing the distribution of the sequence, accurate inference can be achieved for sequences of varying lengths. Experimental results demonstrate the effectiveness of the proposed approach in sleep stage detection and symbol detection in digital communications.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2022)
Article
Automation & Control Systems
Le Yang, Shiji Song, Shuang Li, Yiming Chen, Gao Huang
Summary: The proposed GDR-ELM, a graph embedding-based DR framework, reconstructs all samples according to the weights in a graph matrix containing supervised information, instead of self-reconstruction. GDR-ELM can be stacked as building blocks to construct a multilayer framework for more complicated representation learning tasks, and experiments on various datasets demonstrate its effectiveness.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2021)
Article
Chemistry, Multidisciplinary
Joonyoung Kim, Donghyeon Lee, Kyomin Jung
Summary: Crowdsourcing systems are important for collecting large amounts of qualified data, and K-approval voting is a method that encourages workers to select the top-K alternatives. This paper proposes an efficient algorithm for inferring the correct answers in K-approval voting, which outperforms existing algorithms in extensive experiments.
APPLIED SCIENCES-BASEL
(2021)
Article
Statistics & Probability
Cencheng Shen, Sambit Panda, Joshua T. Vogelstein
Summary: Distance correlation is a popular topic in data science, with a straightforward sample statistic that is ideal for discovering dependency structures. However, the testing process presents a bottleneck due to the need for costly permutation tests. To address this challenge, a chi-squared test method is proposed, which is nonparametric, fast, and applicable to various types of metrics, showing similar testing power to permutation tests.
JOURNAL OF COMPUTATIONAL AND GRAPHICAL STATISTICS
(2022)
Letter
Biochemical Research Methods
Vikram Chandrashekhar, Daniel J. Tward, Devin Crowley, Ailey K. Crow, Matthew A. Wright, Brian Y. Hsueh, Felicity Gore, Timothy A. Machado, Audrey Branch, Jared S. Rosenblum, Karl Deisseroth, Joshua T. Vogelstein
Article
Statistics & Probability
Jesus Arroyo, Elizaveta Levina
Summary: The study focuses on estimating overlapping community memberships in a network, particularly sparse node membership vectors. An algorithm based on sparse principal subspace estimation is developed, showing computational efficiency and eliminating the need for additional clustering steps. The method's fixed point corresponds to correct node memberships in a version of the stochastic block model, with good statistical performance and computational efficiency demonstrated on simulated and real-world networks.
SANKHYA-SERIES A-MATHEMATICAL STATISTICS AND PROBABILITY
(2022)
Editorial Material
Computer Science, Interdisciplinary Applications
Jean-Baptiste Poline, David N. Kennedy, Friedrich T. Sommer, Giorgio A. Ascoli, David C. Van Essen, Adam R. Ferguson, Jeffrey S. Grethe, Michael J. Hawrylycz, Paul M. Thompson, Russell A. Poldrack, Satrajit S. Ghosh, David B. Keator, Thomas L. Athey, Joshua T. Vogelstein, Helen S. Mayberg, Maryann E. Martone
Summary: This perspective article discusses the importance of international collaboration and organizations in promoting standardization of neuroscience data to make it more Findable, Accessible, Interoperable, and Reusable (FAIR). The article emphasizes the current inadequacy of standards as a major obstacle to the interoperability and reusability of research results, calling for increased international collaboration to address this issue.
Article
Statistics & Probability
Jaewon Chung, Bijan Varjavand, Jesus Arroyo-Relion, Anton Alyakin, Joshua Agterberg, Minh Tang, Carey E. Priebe, Joshua T. Vogelstein
Summary: This study focuses on testing whether two graphs come from the same distribution in brain network analysis. By utilizing methods such as adjacency spectral embedding and nonparametric maximum mean discrepancy test, the alignment and distribution comparison issues between different graphs are addressed. The results indicate that using multiscale graph correlation test can lead to more powerful outcomes in Drosophila brain network data.
Article
Medicine, General & Internal
Shilong Li, Tomi Jun, Jonathan Tyler, Emilio Schadt, Yu-Han Kao, Zichen Wang, Maximilian F. Konig, Chetan Bettegowda, Joshua T. Vogelstein, Nickolas Papadopoulos, Ramon E. Parsons, Rong Chen, Eric E. Schadt, Li Li, William K. Oh
Summary: Alpha-1-adrenergic receptor antagonists may improve in-hospital mortality among COVID-19 patients by suppressing pro-inflammatory cytokines. This study found that exposure to alpha-1-blockers was independently associated with improved in-hospital mortality in COVID-19 patients.
FRONTIERS IN MEDICINE
(2022)
Article
Biology
Thomas L. Athey, Daniel J. Tward, Ulrich Mueller, Joshua T. Vogelstein, Michael I. Miller
Summary: ViterBrain is an automated probabilistic reconstruction method that reconstructs neuronal geometry and processes from microscopy images. It combines a hidden Markov state process and a random field appearance model of neuron fluorescence, and uses dynamic programming to compute the most probable neuron path. The method can handle imperfect image segmentations and noise, and provides an interactive framework for users to trace neurons.
COMMUNICATIONS BIOLOGY
(2022)
Article
Neurosciences
Benjamin D. D. Pedigo, Michael Winding, Carey E. E. Priebe, Joshua T. T. Vogelstein
Summary: Graph matching algorithms aim to find the best correspondence between nodes in two networks. They have been used to match individual neurons in nanoscale connectomes, especially those across hemispheres. However, they usually only utilize the same hemisphere subgraphs. We propose a modification to a state-of-the-art graph matching algorithm that allows it to solve the bisected graph matching problem, using connections between brain hemispheres for better neuron pair predictions. We demonstrate improved matching accuracy by combining our approach with previous extensions to graph matching, including edge types and known neuron pairings.
NETWORK NEUROSCIENCE
(2023)
Article
Computer Science, Artificial Intelligence
Hayden S. Helm, Amitabh Basu, Avanti Athreya, Youngser Park, Joshua T. Vogelstein, Carey E. Priebe, Michael Winding, Marta Zlatic, Albert Cardona, Patrick Bourke, Jonathan Larson, Marah Abdin, Piali Choudhury, Weiwei Yang, Christopher W. White
Summary: Learning to rank is a problem of general interest, and we propose an effective approach using an integer linear program. We demonstrate its effectiveness through theoretical analysis, simulations, and real data examples.
PATTERN RECOGNITION
(2023)
Article
Medicine, General & Internal
Mike Powell, Callahan Clark, Anton Alyakin, Joshua T. Vogelstein, Brian Hart
Summary: This retrospective cohort study examines the use of metformin in patients with diabetes and proposes a more robust study design to address residual confounding. The results suggest that common observational study designs for metformin research are at risk for significant confounding.
Article
Biochemical Research Methods
Ting Xu, Gregory Kiar, Jae Wook Cho, Eric W. Bridgeford, Aki Nikolaidis, Joshua T. Vogelstein, Michael P. Milham
Summary: This article presents an integrative toolbox called Reliability eXplorer (ReX) that aids in examining individual variation and reliability in neuroscience biomarker discovery. Additionally, the article introduces a two-dimensional field map-based approach called gradient flows, which is implemented in ReX to identify and represent the most effective optimization direction for measuring individual differences.
Article
Biology
Benjamin D. Pedigo, Mike Powell, Eric W. Bridgeford, Michael Winding, Carey E. Priebe, Joshua T. Vogelstein, Srdjan Ostojic
Summary: Comparing connectomes can provide insights into the relationship between neural connectivity, genetics, disease, development, learning, and behavior. However, statistical inference about differences between two networks is challenging, especially at the nanoscale level. In this study, we investigate the bilateral symmetry of a larval Drosophila brain connectome to refine our understanding of symmetry. We find significant differences in connection probabilities across the left and right networks, as well as between specific cell types. This work demonstrates how statistical inferences from networks can enhance the study of connectomes.
Review
Computer Science, Artificial Intelligence
Itzy Morales E. Pantoja, Lena Smirnova, Alysson R. Muotri, Karl J. Wahlin, Jeffrey Kahn, J. Lomax Boyd, David H. Gracias, Timothy D. Harris, Tzahi Cohen-Karni, Brain S. Caffo, Alexander S. Szalay, Fang Han, Donald J. Zack, Ralph Etienne-Cummings, Akwasi Akwaboah, July Carolina Romero, Dowlette-Mary Alam El Din, Jesse D. Plotkin, Barton L. Paulhamus, Erik C. Johnson, Frederic Gilbert, J. Lowry Curley, Ben Cappiello, Jens C. Schwamborn, Eric J. Hill, Paul Roach, Daniel Tornero, Caroline Krall, Rheinallt Parri, Fenna Sille, Andre Levchenko, Rabih E. Jabbour, Brett J. Kagan, Cynthia A. Berlinicke, Qi Huang, Alexandra Maertens, Kathrin Herrmann, Katya Tsaioun, Raha Dastgheyb, Christa Whelan Habela, Joshua T. Vogelstein, Thomas Hartung
Summary: The brain is a powerful computation system with efficient processing and adaptability. Advances in stem cell technology have led to the development of three-dimensional brain organoids that better mimic human brain functionality. Organoid Intelligence aims to use these capabilities for biocomputing and synthetic intelligence. Understanding how learning changes organoid connectivity can shed light on cognition in the human brain.
FRONTIERS IN ARTIFICIAL INTELLIGENCE
(2023)
Article
Mathematics, Applied
Adam Li, Ronan Perry, Chester Huynh, Tyler M. Tomita, Ronak Mehta, Jesus Arroyo, Jesse Patsolic, Ben Falk, Sridevi Sarma, Joshua Vogelstein
Summary: Decision forests, such as random forests and gradient boosting trees, are highly accurate in unstructured tabular data while convolutional deep networks (ConvNets) outperform forests in structured data like images and time-series. This is partly because networks consider feature indices while naive forest implementations do not. The Manifold Oblique Random Forests (Morf) approach incorporates feature locality by choosing distributions in a manifold-aware manner, achieving fast computation and maintaining interpretability and theoretical justification. Morf exhibits excellent performance in classification tasks on both simulated data and real images and time-series.
SIAM JOURNAL ON MATHEMATICS OF DATA SCIENCE
(2023)
Review
Computer Science, Artificial Intelligence
Dhireesha Kudithipudi, Mario Aguilar-Simon, Jonathan Babb, Maxim Bazhenov, Douglas Blackiston, Josh Bongard, Andrew P. Brna, Suraj Chakravarthi Raja, Nick Cheney, Jeff Clune, Anurag Daram, Stefano Fusi, Peter Helfer, Leslie Kay, Nicholas Ketz, Zsolt Kira, Soheil Kolouri, Jeffrey L. Krichmar, Sam Kriegman, Michael Levin, Sandeep Madireddy, Santosh Manicka, Ali Marjaninejad, Bruce McNaughton, Risto Miikkulainen, Zaneta Navratilova, Tej Pandit, Alice Parker, Praveen K. Pilly, Sebastian Risi, Terrence J. Sejnowski, Andrea Soltoggio, Nicholas Soures, Andreas S. Tolias, Dario Urbina-Melendez, Francisco J. Valero-Cuevas, Gido M. van de Ven, Joshua T. Vogelstein, Felix Wang, Ron Weiss, Angel Yanguas-Gil, Xinyun Zou, Hava Siegelmann
Summary: Biological organisms learn from interactions with their environment, and artificial systems also need the ability to learn throughout their lifetime. This article introduces biological mechanisms and artificial models and mechanisms for lifelong learning, and discusses opportunities to bridge the gap between natural and artificial intelligence.
NATURE MACHINE INTELLIGENCE
(2022)