Article
Chemistry, Physical
Sajal Kumar Giri, George C. Schatz
Summary: This study reports the manipulation of electronic excitations using the unique temporal and spectral features of pulsed entangled photons. A comprehensive optimization protocol based on Bayesian optimization is developed to selectively excite electronic states. The results show that entangled light significantly enhances two-photon absorption probability and enables selective excitation.
JOURNAL OF PHYSICAL CHEMISTRY LETTERS
(2022)
Article
Multidisciplinary Sciences
Tao Yan, Rui Yang, Ziyang Zheng, Xing Lin, Hongkai Xiong, Qionghai Dai
Summary: Photonic neural networks use photons instead of electrons to perform brain-like computations, leading to significantly improved computing performance. However, current architectures are limited to handling data with regular structures and cannot generalize to graph-structured data beyond Euclidean space. In this study, a diffractive graph neural network (DGNN) is proposed to address this limitation by utilizing diffractive photonic computing units (DPUs) and on-chip optical devices. DGNN achieves complex feature representation by capturing dependencies among node neighborhoods during light-speed optical message passing over graph structures. It demonstrates superior performance in node and graph-level classification tasks with benchmark databases, providing a new direction for high-efficiency processing of large-scale graph data structures using deep learning.
Article
Chemistry, Physical
Kanishka Singh, Jannes Munchmeyer, Leon Weber, Ulf Leser, Annika Bande
Summary: In this study, five different Graph Neural Networks (GNNs) were benchmarked and analyzed for the prediction of excitation spectra in organic molecules. The performance of GNNs was compared in terms of runtime measurements, prediction accuracy, and analysis of outliers in the test set. Through TMAP clustering and statistical analysis, clear hotspots of high prediction errors and optimal spectra prediction for molecules with specific functional groups were identified. This in-depth benchmarking and subsequent analysis protocol provides a recipe for comparing different machine learning methods and evaluating dataset quality.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2022)
Article
Biochemical Research Methods
Oezlem Muslu, Charles Tapley Hoyt, Mauricio Lacerda, Martin Hofmann-Apitius, Holger Froehlich
Summary: The study proposes a novel approach, GuiltyTargets, for prioritization of putative targets using attributed network representation learning and positive-unlabeled learning. The evaluation on multiple disease datasets demonstrates its superiority over previous methods and its potential for target repositioning across related diseases.
IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS
(2022)
Article
Computer Science, Artificial Intelligence
Alejandro Moran, Vincent Canals, Fabio Galan-Prado, Christian F. Frasser, Dhinakar Radhakrishnan, Saeid Safavi, Josep L. Rossello
Summary: Edge artificial intelligence is a growing research field, and reservoir computing has attracted attention as a feasible alternative for edge intelligence. This study proposes a simple hardware-optimized circuit design for low-power edge intelligence applications and demonstrates its implementation in FPGA for low-power audio event detection. The results show significant accuracy and ultra-low energy consumption for the proposed approach.
COGNITIVE COMPUTATION
(2023)
Article
Computer Science, Information Systems
Haozhe Chen, Hang Zhou, Jie Zhang, Dongdong Chen, Weiming Zhang, Kejiang Chen, Gang Hua, Nenghai Yu
Summary: In recent years, various methods for protecting model intellectual property (IP) have been proposed, but the problem of quickly detecting copied models among a large number of models on the Internet has not received enough attention. This article introduces a novel model copy detection mechanism called perceptual hashing for convolutional neural networks (CNNs), which can efficiently retrieve similar versions of a query model by comparing hash codes. The experiment demonstrates the superior copy detection performance of the proposed perceptual hashing scheme on a model library of 3,565 models.
ACM TRANSACTIONS ON MULTIMEDIA COMPUTING COMMUNICATIONS AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Leander Weber, Sebastian Lapuschkin, Alexander Binder, Wojciech Samek
Summary: Explainable Artificial Intelligence (XAI) is a research field that aims to bring transparency to complex and opaque machine learning models. This paper provides an overview of techniques that practically apply XAI to improve ML models, categorizing and comparing their strengths and weaknesses. Theoretical perspectives and empirical experiments demonstrate how explanations can enhance properties such as model generalization and reasoning. The potential caveats and drawbacks of these methods are also discussed.
INFORMATION FUSION
(2023)
Article
Business
Baisheng Shi, Hao Wang
Summary: The rapid development of the social economy has significantly impacted the traditional advertising industry, leading to an urgent need for improved accuracy in advertisement promotion and identification. The application of artificial intelligence (AI) methods, particularly the network model based on the genetic algorithm back propagation (GABP) neural network, is highly compatible with the advertising industry. This study applies the GABP neural network in the construction of social networks to predict website advertising click-through rates (CTR) through optimized advertising promotion strategies, resulting in improved accuracy and precision.
TECHNOLOGICAL FORECASTING AND SOCIAL CHANGE
(2023)
Review
Nutrition & Dietetics
Jaroslaw Sak, Magdalena Suchodolska
Summary: Artificial intelligence is increasingly being utilized in biomedical, clinical, and nutritional epidemiology research in the fields of medicine and nutrition science. Different AI methods, such as artificial neural networks, machine learning algorithms, and deep learning algorithms, are applied in various aspects of nutrient studies. AI technology has the potential to revolutionize personalized nutrient supply and monitoring through the development of dietary systems.
Review
Dentistry, Oral Surgery & Medicine
Anita Aminoshariae, Jim Kulild, Venkateshbabu Nagendrababu
Summary: Artificial intelligence in endodontics shows accuracy and precision in detection, determination, and disease prediction, with potential future applications such as robotic-assisted surgery. However, before transferring AI models into clinical practice, further verification of reliability and cost-effectiveness is necessary.
JOURNAL OF ENDODONTICS
(2021)
Article
Chemistry, Physical
Kanishka Singh, Jannes Munchmeyer, Leon Weber, Ulf Leser, Annika Bande
Summary: This work benchmarks and analyzes five different GNNs for predicting excitation spectra of organic molecules, comparing their performance in terms of runtime measurements, prediction accuracy, and analysis of outliers in the test set. Through TMAP clustering and statistical analysis, it identifies hotspots of high prediction errors and optimal spectra prediction for molecules with certain functional groups. This benchmarking and analysis protocol provides a recipe for comparing different ML methods and evaluating dataset quality.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2022)
Article
Optics
Jingxi Li, Yi-Chun Hung, Onur Kulce, Deniz Mengu, Aydogan Ozcan
Summary: This study introduces a polarization-multiplexed diffractive processor that can all-optically perform multiple arbitrarily-selected linear transformations through a single diffractive network trained using deep learning. The transmission layers of this processor are trained and optimized via deep learning to successfully approximate and implement a group of arbitrarily-selected target transformations.
LIGHT-SCIENCE & APPLICATIONS
(2022)
Review
Dentistry, Oral Surgery & Medicine
Fahad Umer, Saqib Habib
Summary: This scoping review analyzed the use of artificial intelligence (AI) algorithms and models in endodontics. Most studies used neural networks, particularly convolutional neural networks. The AI models showed acceptable performance, achieving accuracy rates above 90% in diagnostic tasks. However, there were irregularities in the reporting of AI-related research. The endodontic community should implement recommended guidelines to improve the scientific rigor of AI research.
JOURNAL OF ENDODONTICS
(2022)
Review
Pharmacology & Pharmacy
Shan Wang, Jinwei Di, Dan Wang, Xudong Dai, Yabing Hua, Xiang Gao, Aiping Zheng, Jing Gao
Summary: The rapid development of artificial neural networks in the field of pharmaceutical formulation allows them to replace hundreds of trial and error experiments and become an important method in pharmaceutical science research.
Review
Construction & Building Technology
Giovanni Calzolari, Wei Liu
Summary: Fast and accurate airflow simulations in the built environment are crucial for providing comfortable thermal conditions and air quality. Computational Fluid Dynamics (CFD) offers detailed analysis, but faces challenges in terms of computational cost and accuracy. Deep learning and Artificial Neural Networks (ANN) are increasingly explored as alternative methods to improve simulation efficiency and accuracy.
BUILDING AND ENVIRONMENT
(2021)
Article
Statistics & Probability
Tuomas Sivula, Mans Magnusson, Aki Vehtari
Summary: When using leave-one-out cross-validation to evaluate models, the variance of the sampling distribution is commonly used to assess the uncertainty of the estimate. Although previous studies have shown that no general unbiased variance estimator can be constructed, we demonstrate the possibility of constructing an unbiased estimator considering a specific predictive performance measure and model. The example illustrates the potential to obtain improved, problem-specific, unbiased estimators for assessing the uncertainty in LOO-CV estimation.
COMMUNICATIONS IN STATISTICS-THEORY AND METHODS
(2023)
Article
Statistics & Probability
Federico Pavone, Juho Piironen, Paul-Christian Burkner, Aki Vehtari
Summary: This paper discusses the advantages of using a reference model in variable selection and demonstrates how it can reduce variability and improve stability in the model selection process. It shows that not only does the reference model enhance the performance of projection predictive variables, but it also greatly improves other variable selection methods.
COMPUTATIONAL STATISTICS
(2023)
Article
Chemistry, Medicinal
Lincan Fang, Xiaomi Guo, Milica Todorovic, Patrick Rinke, Xi Chen
Summary: Finding low-energy conformers of organic molecules on nanoclusters is a complex problem, which is further complicated by the constraints imposed by the presence of the cluster and other surrounding molecules. In this study, we modified our active learning molecular conformer search method to address this challenge, particularly focusing on avoiding steric clashes between the molecule and the cluster. Using a cysteine molecule on a gold-thiolate cluster as a model system, we demonstrated that the conformers in the cluster inherited the hydrogen bond types from isolated conformers but exhibited reordered energy rankings and spacings.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2023)
Article
Nanoscience & Nanotechnology
Anastasia Matuhina, G. Krishnamurthy Grandhi, Fang Pan, Maning Liu, Harri Ali-Loytty, Hussein M. Ayedh, Antti Tukiainen, Jan-Henrik Smatt, Vile Vahanissi, Hele Savin, Jingrui Li, Patrick Rinke, Paola Vivo
Summary: In this study, CsMnCl3 nanocrystals (NCs) were synthesized in two polymorphic structures, cubic (c-CsMnCl3) and rhombohedral (r-CsMnCl3), and it was found that c-CsMnCl3 NCs were nonemissive while r-CsMnCl3 NCs exhibited red emission. The results highlight the importance of NC structures in determining their luminescence properties.
ACS APPLIED NANO MATERIALS
(2023)
Article
Engineering, Electrical & Electronic
Muralikrishnan Srinivasan, Jinxiang Song, Alexander Grabowski, Krzysztof Szczerba, Holger K. Iversen, Mikkel N. Schmidt, Darko Zibar, Jochen Schroder, Anders Larsson, Christian Hager, Henk Wymeersch
Summary: Optical interconnects (OIs) based on vertical-cavity surface-emitting lasers (VCSELs) are widely used in data centers, supercomputers, and vehicles for low-cost, high-rate connectivity. Machine learning (ML) techniques, including deep neural networks, have been applied to improve the performance of VCSEL-based OIs. End-to-end (E2E) autoencoder approaches can optimize the entire parameterized transmitters and receivers for ultimate performance. This tutorial paper provides an overview of ML for VCSEL-based OIs, focusing on E2E approaches and addressing the unique challenges of VCSELs.
JOURNAL OF LIGHTWAVE TECHNOLOGY
(2023)
Article
Chemistry, Physical
Daniel Sorvisto, Patrick Rinke, Tuomas P. Rossi
Summary: In this study, the plasmonic hot-carrier generation in noble metal nanoparticles (Ag, Au, and Cu) with single-atom dopants (Ag, Au, Cu, Pd, and Pt) was investigated using first-principles time-dependent density functional theory calculations. The results showed that the dopant element significantly altered the local hot-carrier generation at the dopant atom, while the plasmonic response of the nanoparticle as a whole was not significantly affected. The hot holes at the dopant atom originated from the discrete d-electron states of the dopant, and the energies of these d electron states and hence those of the hot holes depended on the dopant element, suggesting the possibility of tuning hot-carrier generation with suitable dopants.
JOURNAL OF PHYSICAL CHEMISTRY C
(2023)
Article
Multidisciplinary Sciences
Peter Mikula, Oldrich Tomasek, Dusan Romportl, Timothy K. Aikins, Jorge E. Avendano, Bukola D. A. Braimoh-Azaki, Adams Chaskda, Will Cresswell, Susan J. Cunningham, Svein Dale, Gabriela R. Favoretto, Kelvin S. Floyd, Hayley Glover, Tomas Grim, Dominic A. W. Henry, Tomas Holmern, Martin Hromada, Soladoye B. Iwajomo, Amanda Lilleyman, Flora J. Magige, Rowan O. Martin, Marina de A. Maximiano, Eric D. Nana, Emmanuel Ncube, Henry Ndaimani, Emma Nelson, Johann H. van Niekerk, Carina Pienaar, Augusto J. Piratelli, Penny Pistorius, Anna Radkovic, Chevonne Reynolds, Eivin Roskaft, Griffin K. Shanungu, Paulo R. Siqueira, Tawanda Tarakini, Nattaly Tejeiro-Mahecha, Michelle L. Thompson, Wanyoike Wamiti, Mark Wilson, Donovan R. C. Tye, Nicholas D. Tye, Aki Vehtari, Piotr Tryjanowski, Michael A. Weston, Daniel T. Blumstein, Tomas Albrecht
Summary: This study investigates the factors influencing avian tolerance towards humans in open tropical ecosystems. It finds that rural bird populations and those exposed to lower human disturbance have lower tolerance, while larger species with larger clutches and enhanced flight ability are also less tolerant. The study also shows that escape distances increase during the wet season and from longer starting distances.
NATURE COMMUNICATIONS
(2023)
Article
Automation & Control Systems
David Frich Hansen, Tommy Sonne Alstrom, Mikkel N. Schmidt
Summary: In this study, we address the issue of baseline interference in vibrational spectroscopy by using techniques such as surface enhanced Raman spectroscopy (SERS) and near-infrared (NIR) spectroscopy. By applying a high-capacity non-stationary Gaussian process on the baseline, we are able to estimate the signal and baseline simultaneously, resulting in accurate peak parameter estimation and meaningful uncertainty estimates.
CHEMOMETRICS AND INTELLIGENT LABORATORY SYSTEMS
(2023)
Article
Multidisciplinary Sciences
Vitus Besel, Milica Todorovic, Theo Kurten, Patrick Rinke, Hanna Vehkamaki
Summary: In this study, the GeckoQ dataset was created, which includes atomic structures of 31,637 atmospherically relevant molecules. This dataset can accelerate the research on key atmospheric processes driven by low-volatile organic compounds (LVOCs), such as new particle formation and growth. Machine learning tools were used to explore the relationship between structural and thermodynamic properties, and a first application of Gaussian process regression was demonstrated.
Proceedings Paper
Acoustics
Anders S. Olsen, Emil Ortvald, Kristoffer H. Madsen, Mikkel N. Schmidt, Morten Morup
Summary: The development of suitable models for dynamic functional connectivity is crucial for a better understanding of the brain's activity during rest and tasks. This study introduces mixture models and Hidden Markov models that consider the sign-symmetric distribution of eigenvectors on a hypersphere, and demonstrates their performance on synthetic and task-fMRI data.
2023 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH, AND SIGNAL PROCESSING WORKSHOPS, ICASSPW
(2023)
Article
Mathematics, Interdisciplinary Applications
Philip Greengard, Jeremy Hoskins, Charles C. Margossian, Jonah Gabry, Andrew Gelman, Aki Vehtari
Summary: We propose algorithms for evaluating posterior moments of certain Bayesian linear regression models. These algorithms involve analytical marginalization of regression coefficients followed by numerical integration of the remaining low-dimensional density. Our approach drastically reduces run times compared to state-of-the-art MCMC algorithms and overcomes the difficulty of tuning when applied to hierarchical models.
Article
Chemistry, Analytical
Bo Li, Giulia Zappala, Elodie Dumont, Anja Boisen, Tomas Rindzevicius, Mikkel N. Schmidt, Tommy S. Alstrom
Summary: Rapid and accurate detection and quantification of nitroaromatic explosives is crucial for public health and security. We propose an attention-based vision transformer neural network for nitroaromatic explosives' detection and quantification, which outperforms or is on par with existing methods in terms of detection and concentration prediction accuracy.
Proceedings Paper
Computer Science, Artificial Intelligence
Bo Li, Mikkel N. Schmidt, Tommy S. Alstrom, Sebastian U. Stich
Summary: This paper investigates the impact of data heterogeneity on deep neural networks in federated learning and proposes a method to correct model drift on the final layers, which outperforms existing benchmarks with lower communication cost.
2023 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR
(2023)
Article
Chemistry, Physical
Jonas Busk, Mikkel N. Schmidt, Ole Winther, Tejs Vegge, Peter Bjorn Jorgensen
Summary: This research presents a complete framework for training and recalibrating graph neural network ensemble models to accurately predict energy and forces with calibrated uncertainty estimates. The method is demonstrated and evaluated on challenging datasets, achieving good prediction accuracy and uncertainty calibration.
PHYSICAL CHEMISTRY CHEMICAL PHYSICS
(2023)