A5100409331- Machine Learning in Materials Science
- Computational Drug Discovery Methods
- Protein Structure and Dynamics
- X-ray Diffraction in Crystallography
- Generative Adversarial Networks and Image Synthesis
- Computer Graphics and Visualization Techniques
- Forecasting Techniques and Applications
- Various Chemistry Research Topics
- Nanopore and Nanochannel Transport Studies
- Underwater Vehicles and Communication Systems
- Metal-Organic Frameworks: Synthesis and Applications
- Microfluidic and Capillary Electrophoresis Applications
- Model Reduction and Neural Networks
- Underwater Acoustics Research
- Advanced Graph Neural Networks
- CO2 Reduction Techniques and Catalysts
- Receptor Mechanisms and Signaling
- Carbon dioxide utilization in catalysis
- Scientific Research and Discoveries
- Photoreceptor and optogenetics research
- vaccines and immunoinformatics approaches
- Electrostatics and Colloid Interactions
- Target Tracking and Data Fusion in Sensor Networks
- Bacteriophages and microbial interactions
- Brain Tumor Detection and Classification
Materials Research Center
2025
Carnegie Mellon University
2020-2023
Wuhan Ship Development & Design Institute
2023
Harbin Engineering University
2023
Fuzhou University
2023
Yale University
2022
Westlake University
2022
Columbia University
2022
Zhejiang University
2022
Amazon (United States)
2020-2021
Abstract Accurate and efficient prediction of polymer properties is great significance in design. Conventionally, expensive time-consuming experiments or simulations are required to evaluate functions. Recently, Transformer models, equipped with self-attention mechanisms, have exhibited superior performance natural language processing. However, such methods not been investigated sciences. Herein, we report TransPolymer, a Transformer-based model for property prediction. Our proposed...
Metal-organic frameworks (MOFs) are materials with a high degree of porosity that can be used for many applications. However, the chemical space MOFs is enormous due to large variety possible combinations building blocks and topology. Discovering optimal specific applications requires an efficient accurate search over countless potential candidates. Previous high-throughput screening methods using computational simulations like DFT time-consuming. Such also require 3D atomic structures MOFs,...
Abstract Two-dimensional nanomaterials, such as graphene, have been extensively studied because of their outstanding physical properties. Structure and topology nanopores on materials can be important for performances in real-world engineering applications, like water desalination. However, discovering the most efficient often involves a very large number experiments or simulations that are expensive time-consuming. In this work, we propose data-driven artificial intelligence (AI) framework...
Abstract Machine learning (ML) models have been widely successful in the prediction of material properties. However, large labeled datasets required for training accurate ML are elusive and computationally expensive to generate. Recent advances Self-Supervised Learning (SSL) frameworks capable on unlabeled data mitigate this problem demonstrate superior performance computer vision natural language processing. Drawing inspiration from developments SSL, we introduce Crystal Twins (CT): a...
Ionic liquids (ILs) provide a promising solution for CO2 capture and storage to mitigate global warming. However, identifying designing the high-capacity IL from giant chemical space require expensive exhaustive simulations experiments. Machine learning (ML) can accelerate process of searching desirable ionic molecules through accurate efficient property predictions in data-driven manner. existing descriptors ML models molecule suffer inefficient adaptation molecular graph structure....
Recent advances in equivariant graph neural networks (GNNs) have made deep learning amenable to developing fast surrogate models expensive
Abstract The current design of aerodynamic shapes, like airfoils, involves computationally intensive simulations to explore the possible space. Usually, such relies on prior definition parameters and places restrictions synthesizing novel shapes. In this work, we propose a data-driven shape encoding generating method, which automatically learns representations from existing airfoils uses learned generate new airfoils. are then used in optimization synthesized airfoil shapes based their...
Abstract Machine learning (ML) has demonstrated the promise for accurate and efficient property prediction of molecules crystalline materials. To develop highly ML models chemical structure prediction, datasets with sufficient samples are required. However, obtaining clean data properties can be expensive time-consuming, which greatly limits performance models. Inspired by success augmentations in computer vision natural language processing, we developed AugLiChem: augmentation library...
GPCRs are the target for one-third of FDA-approved drugs, however; development new drug molecules targeting is limited by lack mechanistic understanding GPCR structure-activity-function relationship. To modulate activity with highly specific drugs and minimal side-effects, it necessary to quantitatively describe important structural features in correlate them activation state GPCR. In this study, we developed 3 ML approaches predict conformation proteins. Additionally, level based on their...
Human perception systems are highly refined, relying on an adaptive, plastic, and event-driven network of sensory neurons. Drawing inspiration from Nature, neuromorphic hold tremendous potential for efficient multisensory signal processing in the physical world; however, development artificial neuron with a widely calibratable spiking range reduced footprint remains challenging. Here, we report organic electrochemical (OECN) (<37 mm 2 ) based high-performance vertical OECT (vOECT)...
Tensor product function (TPF) approximations have been widely adopted in solving high-dimensional problems, such as partial differential equations and eigenvalue achieving desirable accuracy with computational overhead that scales linearly problem dimensions. However, recent studies underscored the extraordinarily high cost of TPFs on quantum many-body even for systems few three particles. A key distinction these problems is antisymmetry requirement unknown functions. In present work, we...
Underwater acoustic target recognition methods based on time-frequency analysis have shortcomings, such as missing information characteristics and having a large computation volume, which leads to difficulties in improving the accuracy immediacy of system. In this paper, an underwater model deep residual attention convolutional neural network called DRACNN is proposed, whose input time-domain signal targets radiated noise. model, blocks with mechanisms are used focus extract features target,...
Abstract The majority of computational catalyst design focuses on the screening material components and alloy composition to optimize selectivity activity for a given reaction. However, predicting metastability surface at realistic operating conditions requires an extensive sampling possible reconstructions their associated kinetic pathways. We present CatGym, deep reinforcement learning (DRL) environment thermal reconstruction pathways barriers in crystalline solids under reaction...
Intermittency are a common and challenging problem in demand forecasting. We introduce new, unified framework for building probabilistic forecasting models intermittent time series, which incorporates allows to generalize existing methods several directions. Our is based on extensions of well-established model-based discrete-time renewal processes, can parsimoniously account patterns such as aging, clustering quasi-periodicity arrivals. The connection processes not only principled extension...
Abstract Design and optimization of hull shapes for optimal hydrodynamic performance have been a major challenge naval architectures. Deep learning bears the promise comprehensive geometric representation new design synthesis. In this work, we develop deep neural network (DNN)-based approach to encode designs condensed representations, synthesize novel designs, optimize synthetic based on performance. A variational autoencoder (VAE) with hydro-predictor is developed learn through...
In this paper we tackle the problem of generating conformers a molecule in 3D space given its molecular graph. We parameterize these as continuous functions that map elements from graph to points space. then formulate learning generate distribution over using diffusion generative model, called Molecular Conformer Fields (MCF). Our approach is simple and scalable, achieves state-of-the-art performance on challenging conformer generation benchmarks while making no assumptions about explicit...