Article
Multidisciplinary Sciences
Kaiwen Wang, Yuqiu Yang, Fangjiang Wu, Bing Song, Xinlei Wang, Tao Wang
Summary: While scRNA-seq data analysis techniques are advanced, research on CyTOF data analysis has lagged behind. Dimension reduction methods were benchmarked on real and synthetic CyTOF samples, highlighting the high level of complementarity between the methods. The study provides useful guidelines for choosing the appropriate method based on data structure and analytical needs.
NATURE COMMUNICATIONS
(2023)
Review
Immunology
Zicheng Hu, Sanchita Bhattacharya, Atul J. Butte
Summary: Modern cytometry technologies allow profiling of the immune system at a single-cell resolution with over 50 protein markers, and the number of publicly available cytometry datasets is increasing. Analyzing cytometry data remains challenging due to its high dimensionality, large cell numbers, and dataset heterogeneity, but machine learning techniques are well suited for addressing these challenges and have been employed in various aspects of cytometry data analysis.
FRONTIERS IN IMMUNOLOGY
(2022)
Article
Automation & Control Systems
Rick van Veen, Michael Biehl, Gert-Jan De Vries
Summary: sklvq is an open-source Python implementation of learning vector quantization algorithms, known for its modular and customizable design. Users can easily extend the algorithms, and detailed documentation and rich API make it easy to use.
JOURNAL OF MACHINE LEARNING RESEARCH
(2021)
Article
Biochemistry & Molecular Biology
Andrei S. Rodin, Grigoriy Gogoshin, Seth Hilliard, Lei Wang, Colt Egelston, Russell C. Rockne, Joseph Chao, Peter P. Lee
Summary: This study focuses on predicting clinical responses in cancer immunotherapy by identifying pretreatment biomarkers and analyzing changes in immune networks before and after treatment. By utilizing a novel computational pipeline and systems biology/machine learning techniques, strong signals that may be missed by conventional methods can be detected. Future studies will aim to validate and track immune biomarkers associated with clinical responses identified using this computational pipeline.
INTERNATIONAL JOURNAL OF MOLECULAR SCIENCES
(2021)
Article
Immunology
Hannah den Braanker, Margot Bongenaar, Erik Lubberts
Summary: Spectral flow cytometry is a technique that allows for multicolor panels and investigation of numerous cellular parameters. This article discusses the challenges of preparing and analyzing spectral flow cytometry data, presenting a workflow for high-dimensional analysis. The provided R-based pipeline aims to aid users in obtaining valid and reproducible results.
FRONTIERS IN IMMUNOLOGY
(2021)
Article
Computer Science, Artificial Intelligence
MengLing Fan, Fengzhen Tang, Yinan Guo, Xingang Zhao
Summary: This paper proposes a novel dynamic generalized learning Riemannian space quantization (DGLRSQ) method, which represents each instance by a sequence of covariance matrices and incorporates a short-term memory mechanism to capture the temporal evolution of correlation in SPD matrix-valued data.
PATTERN RECOGNITION
(2022)
Article
Engineering, Environmental
Ji Won Park, Joby Boxall, Sung Kyu Maeng
Summary: Culture-independent data can be used to identify HPC exceedances in drinking water. The study developed an ANN model using ICC, ATP, and chlorine data, achieving high accuracy in classifying HPC exceedances. The model overcomes culture dependence and provides near real-time data for ensuring the safety of drinking water.
Article
Multidisciplinary Sciences
David Ross
Summary: Flow cytometry is widely used for evaluating engineered bacteria, requiring automated analysis methods. FlowGateNIST, a Python package, offers automatic analysis for bacterial flow cytometry data, including automatic gating and fluorescence signal calibration. This open source tool aims to support users with minimal programming experience.
Article
Biochemical Research Methods
Yuting Dai, Aining Xu, Jianfeng Li, Liang Wu, Shanhe Yu, Jun Chen, Weili Zhao, Xiao-Jian Sun, Jinyan Huang
Summary: CytoTree is a versatile tool for analyzing multidimensional flow and mass cytometry data, providing various computational functionalities and supporting the construction of tree-shaped trajectories. Its practical utility is demonstrated through several examples of mass cytometry and time-course flow cytometry data.
BMC BIOINFORMATICS
(2021)
Article
Computer Science, Artificial Intelligence
Moritz Heusinger, Christoph Raab, Frank-Michael Schleif
Summary: This work modifies robust soft learning vector quantization and generalized learning vector quantization to handle concept drift in streaming data and applies momentum-based stochastic gradient descent techniques. Tested against common benchmark algorithms and streaming data in the field, the proposed work achieved promising results.
NEURAL COMPUTING & APPLICATIONS
(2022)
Article
Biochemical Research Methods
Yiming Wang, Ziwei Huang, Xiaojie Wang, Fengrui Yang, Xuebiao Yao, Tingrui Pan, Baoqing Li, Jiaru Chu
Summary: Fluorescence imaging flow cytometry (IFC) is an important biomedical technique for analyzing specific cell subpopulations. However, motion blur in cell images caused by the high-speed flow of fluorescent cells makes it challenging to identify cell types. In this study, a real-time single-cell imaging and classification system based on a fluorescence microscope and deep learning algorithm was developed to directly identify cell types from motion-blur images. The system achieved a high accuracy of 96.6% for single-cell classification of HeLa cells in three different mitotic stages, with a short processing time of only 2 ms.
Article
Biology
Matthias Woedlinger, Michael Reiter, Lisa Weijler, Margarita Maurer-Granofszky, Angela Schumich, Elisa O. Sajaroff, Stefanie Groeneveld-Krentz, Jorge G. Rossi, Leonid Karawajew, Richard Ratei, Michael N. Dworzak
Summary: This study presents an automated method to compute the Minimal Residual Disease (MRD) value directly from FCM data in Acute Lymphoblastic Leukemia (ALL) patients. The method utilizes a neural network based on the transformer architecture to identify blast cells in samples. Evaluation on multiple datasets shows that the proposed method outperforms existing methods.
COMPUTERS IN BIOLOGY AND MEDICINE
(2022)
Article
Biochemical Research Methods
Kyle Ferchen, Nathan Salomonis, H. Leighton Grimes
Summary: Conventional flow cytometry is limited in its ability to detect markers, while new strategies like Infinity Flow can generate and impute hundreds of markers in millions of cells. This study introduces a Python workflow, pyInfinityFlow, for analyzing Infinity Flow data, allowing for the efficient analysis of millions of cells without down-sampling. The workflow accurately identifies both common and rare cell populations and can nominate novel markers for flow cytometry gating strategies.
Article
Biotechnology & Applied Microbiology
Adam Chan, Wei Jiang, Emily Blyth, Jean Yang, Ellis Patrick
Summary: The study introduces a framework called treekoR, which empirically recapitulates cellular structures and facilitates multiple quantifications and comparisons of cell type proportions. Results from twelve case studies emphasize the importance of quantifying proportions relative to parent populations in cytometry data analysis.
Article
Engineering, Electrical & Electronic
Shih Yu Chang, Hsiao-Chun Wu
Summary: This paper introduces the quantization technique and its application in image compression. A new tensor quantization (TQ) framework is proposed to avoid reducing the dimensionality of image data and destroying the two-dimensional spatial relationship. Experimental results demonstrate the superiority of the TQ approach, especially for high-dimensional images.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2022)
Article
Computer Science, Artificial Intelligence
M. Straat, F. Abadi, Z. Kan, C. Goepfert, B. Hammer, M. Biehl
Summary: A modelling framework is presented for investigating supervised learning in non-stationary environments, focusing on LVQ for classification and neural networks for regression. Standard LVQ algorithms are found to be somewhat suitable for training in non-stationary environments, but weight decay does not improve performance under drift processes. The sensitivity to concept drift and effectiveness of weight decay differ significantly between different activation functions in neural networks.
NEURAL COMPUTING & APPLICATIONS
(2022)
Editorial Material
Computer Science, Artificial Intelligence
Luca Oneto, Kerstin Bunte, Nicolo Navarin
Article
Computer Science, Artificial Intelligence
Abolfazl Taghribi, Marco Canducci, Michele Mastropietro, Sven De Rijcke, Kerstin Bunte, Peter Tino
Summary: This article presents an efficient sub-sampling strategy for preserving homology in high-dimensional data. Additionally, a technique for probabilistic description of significant cycles and cavities in the data is proposed.
Article
Astronomy & Astrophysics
S. Rezaei, J. P. McKean, M. Biehl, A. Javadpour
Summary: DECORAS is a deep-learning-based approach that can detect both point and extended sources from VLBI observations. It provides source characterization in terms of position, effective radius, and peak brightness, and outperforms traditional methods in completeness and purity. It can accurately recover the position, effective radius, and peak surface brightness of the detected sources.
MONTHLY NOTICES OF THE ROYAL ASTRONOMICAL SOCIETY
(2022)
Article
Computer Science, Artificial Intelligence
Mohammad Mohammadi, Peter Tino, Kerstin Bunte
Summary: The presence of manifolds is a common assumption in many applications. This article proposes a new algorithm that utilizes random walkers and local alignment measure to detect and denoise manifolds, inspired by the collective behavior of biological ants. The algorithm shows promising results in theoretical analysis and empirical experiments.
NEURAL COMPUTATION
(2022)
Article
Astronomy & Astrophysics
M. Mohammadi, J. Mutatiina, T. Saifollahi, K. Bunte
Summary: This study aims to separate Ultra-compact dwarfs (UCDs) and Globular Clusters (GCs) from foreground stars and background galaxies using multi-wavelength imaging data. The results show that angular sizes and certain color indices are important markers for this classification problem.
ASTRONOMY AND COMPUTING
(2022)
Article
Astronomy & Astrophysics
S. Rezaei, J. P. McKean, M. Biehl, W. de Roo, A. Lafontaine
Summary: We propose a machine learning based approach to detect galaxy-scale gravitational lenses from interferometric data taken by the International LOFAR Telescope. By training and testing Convolutional Neural Networks on simulated data, we achieve a high true positive rate (95.3%) and a low false positive rate (0.008%). The method shows robustness when maximum image separation is >= 3 times the synthesized beam size, and when the lensed images have a total flux density >= 20 sigma.
MONTHLY NOTICES OF THE ROYAL ASTRONOMICAL SOCIETY
(2022)
Article
Computer Science, Interdisciplinary Applications
Rick van Veen, Sanne K. Meles, Remco J. Renken, Fransje E. Reesink, Wolfgang H. Oertel, Annette Janzen, Gert-Jan De Vries, Klaus L. Leenders, Michael Biehl
Summary: In this study, the machine learning algorithm GMLVQ was applied to classify patients with neurodegenerative disorders using FDG-PET scans. The research also demonstrated the visualization of disease progression in the prodromal stages by projecting the scans of iRBD patients into the GMLVQ space. The results showed a correlation between the speed of progression and the change in motor symptoms.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2022)
Article
Automation & Control Systems
Matteo Marcantoni, Bayu Jayawardhana, Mariano Perez Chaher, Kerstin Bunte
Summary: Recent developments in communication technologies and computing paradigms have provided further possibilities for real-time networked control systems. However, privacy and cyber-security concerns arise when sharing private data. This paper presents a secure version of distributed formation control using fully homomorphic encryption and a mixed quantizer, and analyzes its stability.
IEEE CONTROL SYSTEMS LETTERS
(2022)
Review
Business, Finance
Htet Htet Htun, Michael Biehl, Nicolai Petkov
Summary: The identification of critical features is crucial in stock market forecasting for accurate predictions. This survey analyzes 32 research works that combine feature study and ML approaches in various stock market applications. The most widely used feature selection and extraction techniques for accurate predictions include correlation criteria, random forest, principal component analysis, and autoencoder.
FINANCIAL INNOVATION
(2023)
Article
Astronomy & Astrophysics
Petra Awad, Reynier Peletier, Marco Canducci, Rory Smith, Abolfazl Taghribi, Mohammad Mohammadi, Jihye Shin, Peter Tino, Kerstin Bunte
Summary: The distribution of galaxies and clusters of galaxies on the mega-parsec scale of the Universe, known as the Large-Scale Structure or the Cosmic Web, follows an intricate pattern. The 1-Dimensional Recovery, Extraction, and Analysis of Manifolds (1-dream) toolbox is able to extract structures in the Cosmic Web and create probabilistic models of them. Compared to other methodologies, 1-DREAM is able to split the network into its various environments with comparable results.
MONTHLY NOTICES OF THE ROYAL ASTRONOMICAL SOCIETY
(2023)
Article
Automation & Control Systems
Matteo Marcantoni, Bayu Jayawardhana, Kerstin Bunte
Summary: In this paper, a scalable dynamic bearing estimator is proposed to obtain the relative bearing of static landmarks in real-time. Convergence analysis of the estimator is provided using contraction theory, along with upper and lower bounds for the estimator gain. Numerical simulations demonstrate the effectiveness of the proposed method.
IEEE CONTROL SYSTEMS LETTERS
(2023)
Article
Computer Science, Artificial Intelligence
Abolfazl Taghribi, Kerstin Bunte, Rory Smith, Jihye Shin, Michele Mastropietro, Reynier F. Peletier, Peter Tino
Summary: Dimensionality reduction and clustering are important preprocessing steps for machine learning tasks. However, the presence of noise and outliers can greatly affect their performance. In this study, we propose a novel method based on Ant colony optimization to extract manifolds from noisy data. Our technique captures points aligned with major directions of the manifold, and the use of ant pheromone further enhances this behavior. We demonstrate the algorithm's performance on synthetic and real datasets, including an N-body simulation.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Thomas Villmann, Daniel Staps, Jensun Ravichandran, Sascha Saralajew, Michael Biehl, Marika Kaden
Summary: This method allows for classification using data from multiple sources without explicit transfer learning by utilizing a siamese-like GMLVQ architecture. The architecture includes different sets of prototypes for target classification and source separation learning, and trains a linear map for source distinction in parallel to the classification task learning.
ADVANCES IN INTELLIGENT DATA ANALYSIS XX, IDA 2022
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Michiel Straat, Kevin Koster, Nick Goet, Kerstin Bunte
Summary: Insufficient steel quality in mass production can lead to costly damage and low quality products. This study proposes a non-invasive method to measure material properties in real-time during production and predicts key material properties using a linear model. Through analyzing real production data, the study demonstrates an accurate predictive model and validates the relationship between material properties and product faults.
2022 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN)
(2022)