Article
Nanoscience & Nanotechnology
Yoonbeom Park, Kyoungah Cho, Sangsig Kim
Summary: This study used machine learning to predict the output power of hybrid energy devices (HEDs) consisting of photovoltaic cells (PVCs) and thermoelectric generators (TEGs). It found that different interface materials have an impact on the HED performance, and a carbon paste interface material can increase the output power by 2.6%.
ACS APPLIED MATERIALS & INTERFACES
(2022)
Article
Engineering, Environmental
Haiping Gao, Shifa Zhong, Raghav Dangayach, Yongsheng Chen
Summary: Ultrafiltration (UF) is a widely used membrane-based technology for water and wastewater treatment. This study employed machine learning to establish the correlation between membrane performance indices, membrane properties, and fabrication conditions. The loading of additives and the polymer content were found to be the most significant features affecting membrane performance. Our approach provides practical guidance for the design of separation membranes through data-driven virtual experiments.
ENVIRONMENTAL SCIENCE & TECHNOLOGY
(2023)
Article
Thermodynamics
Qingxiang Ji, Xueyan Chen, Jun Liang, Vincent Laude, Sebastien Guenneau, Guodong Fang, Muamer Kadic
Summary: The study demonstrates a general process to design thermal harvesting devices using optimized composite microstructures with available natural materials. This approach can achieve good thermal energy harvesting performances and mimic the behavior of transformed materials. It provides a beneficial tool to explore other transformation optics enabled designs.
INTERNATIONAL JOURNAL OF HEAT AND MASS TRANSFER
(2021)
Article
Computer Science, Information Systems
Xing Chen, Jingtao Li, Chaitali Chakrabarti
Summary: SplitFed learning is a decentralized learning framework for IoT devices that preserves data privacy, but it has high communication overhead. To reduce this overhead, a selective model update method based on energy and loss changes is proposed, which can save energy while maintaining model accuracy.
JOURNAL OF SIGNAL PROCESSING SYSTEMS FOR SIGNAL IMAGE AND VIDEO TECHNOLOGY
(2022)
Editorial Material
Multidisciplinary Sciences
Ying-Lang Wang, Mao-Chih Huang
Summary: Engineers and algorithms have competed in a virtual test to design a step in the process of manufacturing computer chips. Pairing human expertise with computational efficiency proves most cost-effective, but only when the timing is right.
Article
Chemistry, Physical
Igor Baskin, Yair Ein-Eli
Summary: Electrochemoinformatics, as a scientific discipline, applies information technologies, such as data science, machine learning, and artificial intelligence, to solve problems in electrochemical processes. It is closely related to chemoinformatics and materials informatics and primarily used in battery science and technology.
ADVANCED ENERGY MATERIALS
(2022)
Article
Geosciences, Multidisciplinary
Spyros Kondylatos, Ioannis Prapas, Michele Ronco, Ioannis Papoutsis, Gustau Camps-Valls, Maria Piles, Miguel-Angel Fernandez-Torres, Nuno Carvalhais
Summary: Climate change worsens the occurrence of large wildfires by increasing extreme droughts and heatwaves. This study uses Deep Learning to predict wildfire danger and explainable Artificial Intelligence to analyze model attributions. The presented methodology improves the accuracy of wildfire anticipation and reveals the contribution of different variables.
GEOPHYSICAL RESEARCH LETTERS
(2022)
Article
Biochemical Research Methods
Douglas E. Pires, Keith A. Stubbs, Joshua S. Mylne, David B. Ascher
Summary: Herbicides have played a significant role in weed management, increasing crop yields, and improving food security. However, their widespread use has also led to resistance and environmental concerns. Despite the need for new herbicides with different mechanisms of action, there have been no new herbicides introduced in the market for the past three decades. To address this gap, researchers have developed cropCSM, a computational platform for identifying new, potent, non-toxic, and environmentally safe herbicides. This platform uses knowledge-based approaches and predictive models to guide herbicide design and prioritize screening libraries.
BRIEFINGS IN BIOINFORMATICS
(2022)
Article
Chemistry, Physical
Daniil Yurchenko, Lucas Queiroz Machado, Junlei Wang, Chris Bowen, Suleiman Sharkh, Mohamed Moshrefi-Torbati, Dimitri V. Val
Summary: The paper introduces a novel methodology for developing high-power energy harvesting devices using piezoelectric beams, with a global multidimensional constrained optimization algorithm. By incorporating a specially designed electrical circuit, the device's efficiency can be increased by 35% compared to standard energy harvesting circuits with independent rectifiers.
Article
Agronomy
Humna Khan, Travis J. Esau, Aitazaz A. Farooque, Farhat Abbas
Summary: This study assessed the performance of three machine learning algorithms in predicting wild blueberry harvest losses. The support vector regression model showed the most accurate predictions of ground loss, indicating its usefulness in reducing blueberry losses in the selected fields.
Article
Physics, Applied
Boya Jin, Aaron Brettin, Grant W. Bidney, Nicholaos I. Limberopoulos, Joshua M. Duran, Gamini Ariyawansa, Igor Anisimov, Augustine M. Urbas, Kenneth W. Allen, Sarath D. Gunapala, Vasily N. Astratov
Summary: A design of light-harvesting low-index microconical arrays was proposed to increase the sensitivity and signal-to-noise ratio of MWIR focal plane arrays in thermal cameras. Numerical modeling was used to demonstrate significant power enhancement using slightly tapered microcones, potentially allowing for improved performance of MWIR imaging devices. Experimental results showed a threefold enhancement in photocurrent response, indicating the potential for increased SNR and operation temperature.
APPLIED PHYSICS LETTERS
(2021)
Article
Immunology
Dilraj Kaur, Sumeet Patiyal, Chakit Arora, Ritesh Singh, Gaurav Lodhi, Gajendra P. S. Raghava
Summary: Defensins, host defense peptides present in almost all living species, play a crucial role in innate immunity. By analyzing the differences between defensins and AMPs, certain residues were found to be more abundant in defensins. Machine learning models were successfully developed to predict defensins with high accuracy.
FRONTIERS IN IMMUNOLOGY
(2021)
Article
Computer Science, Artificial Intelligence
Damien Dablain, Kristen N. Jacobson, Colin Bellinger, Mark Roberts, Nitesh V. Chawla
Summary: Convolutional neural networks (CNNs) struggle to generalize to minority classes and have opaque decision-making processes on imbalanced image data. This study focuses on the latent features of CNNs to demystify their decisions. The class top-K CE and FE are found to contain important information regarding a CNN's ability to generalize to minority classes. This research also highlights the significance of diversifying class latent features in developing effective methods for imbalanced learning.
Article
Thermodynamics
Yinan Li, Jun Wang, Chenglong Fu, Liulian Huang, Lihui Chen, Yonghao Ni, Qinghong Zheng
Summary: Green cellulose-based solar light capturing composite films consisting of carbon nanotubes (CNT) and cellulose nanofibrils (CNF) with high photothermal conversion performance were developed. The composite films showed excellent mechanical properties due to the strong electrostatic interactions between CNFs and cationic-modified CNTs. The effect of CNT/CNF ratio on the photothermal conversion capacity and coefficient of thermal expansion (CTE) was investigated.
ENERGY CONVERSION AND MANAGEMENT
(2023)
Article
Materials Science, Multidisciplinary
Amir Zavareh, Brittany Tran, Christian Orred, Savannah Rhodes, Md Saifur Rahman, Myeong Namkoong, Ricky Lee, Cody Carlisle, Miguel Rosas, Anton Pavlov, Ian Chen, Greg Schilling, Marc Smith, Fahad Masood, John Hanks, Limei Tian
Summary: This study presents a soft wearable thermal device that can accurately monitor core body temperature and overcome the limitations of existing monitoring methods.
ADVANCED MATERIALS TECHNOLOGIES
(2023)
Article
Chemistry, Medicinal
Devon P. Holst, Pascal Friederich, Alan Aspuru-Guzik, Timothy P. Bender
Summary: This study used various computational methods to calibrate and compare the frontier orbital energies and optical gaps of novel boron subphthalocyanine derivatives and related compounds. The results showed that computationally inexpensive semiempirical methods outperformed most density functional theory methods for calibration. By using free software and a standard laptop, researchers can confidently determine the physical properties of these materials before the synthesis and purification process.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2022)
Article
Chemistry, Multidisciplinary
Martin Seifrid, Riley J. Hickman, Andres Aguilar-Granda, Cyrille Lavigne, Jenya Vestfrid, Tony C. Wu, Theophile Gaudin, Emily J. Hopkins, Alan Aspuru-Guzik
Summary: Self-driving laboratories, in the form of automated experimentation platforms guided by machine learning algorithms, have emerged as a potential solution to the need for accelerated science. While automated synthesis remains a bottleneck, combining automated and manual synthesis efforts significantly expands the explorable chemical space. Quantifying the cost and considering the capabilities of both automated and manual synthesis can help determine the most efficient synthetic route.
ACS CENTRAL SCIENCE
(2022)
Article
Chemistry, Physical
Phillip W. K. Jensen, Lasse Bjorn Kristensen, Cyrille Lavigne, Alan Aspuru-Guzik
Summary: This study explores the application of molecules and molecular electronics in quantum computing, constructing one-qubit gates using scattering in molecules and two-qubit controlled-phase gates using electron-electron scattering along metallic leads. Furthermore, a class of circuit implementations is proposed, and the framework is demonstrated by illustrating one-qubit gates using the molecular electronic structure of molecular hydrogen as a baseline model.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2022)
Correction
Chemistry, Multidisciplinary
Sungwon Kim, Juhwan Noh, Geun Ho Gu, Alan Aspuru-Guzik, Yousung Jung
ACS CENTRAL SCIENCE
(2022)
Article
Materials Science, Multidisciplinary
Simon Schweidler, Henrik Schopmans, Patrick Reiser, Evgeniy Boltynjuk, Jhon Jairo Olaya, Surya Abhishek Singaraju, Franz Fischer, Horst Hahn, Pascal Friederich, Leonardo Velasco
Summary: High-entropy alloys offer a wide research area for new material compositions and applications. A high-throughput magnetron sputtering synthesis method is presented to fabricate a new HEA gradient layer, allowing for the study of the composition of the HEA system and the influence of individual elements on material properties.
ADVANCED ENGINEERING MATERIALS
(2023)
Review
Nanoscience & Nanotechnology
Zhenpeng Yao, Yanwei Lum, Andrew Johnston, Luis Martin Mejia-Mendoza, Xin Zhou, Yonggang Wen, Alan Aspuru-Guzik, Edward H. Sargent, Zhi Wei Seh
Summary: This Perspective highlights the recent advances in machine learning-driven energy research and proposes a set of key performance indicators to compare the benefits of different ML-accelerated workflows in the field of renewable energy.
NATURE REVIEWS MATERIALS
(2023)
Article
Multidisciplinary Sciences
Pauric Bannigan, Zeqing Bao, Riley J. Hickman, Matteo Aldeghi, Florian Hase, Alan Aspuru-Guzik, Christine Allen
Summary: Long-acting injectables are considered promising for chronic disease treatment, and this study demonstrates the use of machine learning to predict drug release and guide the design of new formulations. The data-driven approach has the potential to reduce development time and cost.
NATURE COMMUNICATIONS
(2023)
Article
Chemistry, Medicinal
Po-Yu Kao, Ya-Chu Yang, Wei-Yin Chiang, Jen-Yueh Hsiao, Yudong Cao, Alex Aliper, Feng Ren, Alan Aspuru-Guzik, Alex Zhavoronkov, Min-Hsiu Hsieh, Yen-Chu Lin
Summary: This article explores the application of hybrid quantum-classical generative adversarial networks (GAN) in drug discovery. By substituting each element of GAN with a variational quantum circuit (VQC), small molecule discovering is achieved. Applying VQC in both the noise generator and discriminator, it can generate small molecules with better physicochemical properties and performance while having fewer trainable parameters. However, the hybrid quantum-classical GANs still face challenges in generating unique and valid molecules compared to their classical counterparts.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2023)
Article
Chemistry, Medicinal
Stanley Lo, Martin Seifrid, Theeophile Gaudin, Alaan Aspuru-Guzik
Summary: One of the biggest challenges in polymer property prediction is finding an effective representation that accurately captures the sequence of repeat units. Inspired by data augmentation techniques in computer vision and natural language processing, we explore rearranging the molecular representation iteratively while preserving connectivity to augment polymer data and reveal additional substructural information. We evaluate the impact of this technique on machine learning models trained on three polymer datasets and compare it to common molecular representations. Data augmentation does not significantly improve machine learning property prediction performance compared to non-augmented representations, except in datasets where the target property is primarily influenced by the polymer sequence.
JOURNAL OF CHEMICAL INFORMATION AND MODELING
(2023)
Article
Chemistry, Physical
Philipp Schleich, Joseph Boen, Lukasz Cincio, Abhinav Anand, Jakob S. Kottmann, Sergei Tretiak, Pavel A. Dub, Alan Aspuru-Guzik
Summary: The limited availability of noisy qubits in current quantum computing hardware restricts the investigation of larger, more complex molecules in quantum chemistry calculations. In this study, a classical and near-classical treatment within the framework of quantum circuits is explored. A product ansatz for the parametrized wavefunction is used, along with post-treatment to account for interactions between subsystems. The circuit structure is molecule-dependent and is constructed using simulated annealing and genetic algorithms.
JOURNAL OF CHEMICAL THEORY AND COMPUTATION
(2023)
Article
Computer Science, Interdisciplinary Applications
Sergio Pablo-Garcia, Santiago Morandi, Rodrigo A. Vargas-Hernandez, Kjell Jorner, Zarko Ivkovic, Nuria Lopez, Alan Aspuru-Guzik
Summary: GAME-Net is a graph deep learning model trained with small molecules containing a wide set of functional groups for predicting the adsorption energy of closed-shell organic molecules on metal surfaces, avoiding expensive density functional theory simulations. The model yields a mean absolute error of 0.18 eV on the test set and is 6 orders of magnitude faster than density functional theory.
NATURE COMPUTATIONAL SCIENCE
(2023)
Article
Chemistry, Multidisciplinary
Naruki Yoshikawa, Kourosh Darvish, Mohammad Ghazi Vakili, Animesh Garg, Alan Aspuru-Guzik
Summary: Self-driving laboratories require robotic liquid handling and transfer, and we propose a 3D-printed digital pipette design that overcomes the limitations of current robot grippers. It is cost-effective and easy to assemble, and performance evaluation shows comparable precision to commercial devices.
Review
Physics, Applied
Mario Krenn, Robert Pollice, Si Yue Guo, Matteo Aldeghi, Alba Cervera-Lierta, Pascal Friederich, Gabriel Dos Passos Gomes, Florian Hase, Adrian Jinich, AkshatKumar Nigam, Zhenpeng Yao, Alan Aspuru-Guzik
Summary: Scientists aim to understand the principles behind predictions rather than just being satisfied with accurate results. With the advancement of computational power and artificial intelligence, the role of these systems in contributing to scientific understanding becomes a significant question.
NATURE REVIEWS PHYSICS
(2022)
Review
Chemistry, Multidisciplinary
Martin Seifrid, Robert Pollice, Andres Aguilar-Granda, Zamyla Morgan Chan, Kazuhiro Hotta, Cher Tian Ser, Jenya Vestfrid, Tony C. Wu, Alan Aspuru-Guzik
Summary: To address climate change and disease risks, it is crucial to accelerate technological advancements through better integration between hypothesis generation, design, experimentation, and data analysis. Automated laboratories can significantly speed up molecular and materials discovery by generating information-rich data. Open high-quality datasets will enhance the accessibility and reproducibility of science. This paper presents successful efforts in building self-driving laboratories for the development of new materials.
ACCOUNTS OF CHEMICAL RESEARCH
(2022)
