skip to main content


Title: Efficient and Equivariant Graph Networks for Predicting Quantum Hamiltonian
We consider the prediction of the Hamiltonian matrix, which finds use in quantum chemistry and condensed matter physics. Efficiency and equiv- ariance are two important, but conflicting factors. In this work, we propose a SE(3)-equivariant net- work, named QHNet, that achieves efficiency and equivariance. Our key advance lies at the inno- vative design of QHNet architecture, which not only obeys the underlying symmetries, but also en- ables the reduction of number of tensor products by 92%. In addition, QHNet prevents the expo- nential growth of channel dimension when more atom types are involved. We perform experiments on MD17 datasets, including four molecular sys- tems. Experimental results show that our QHNet can achieve comparable performance to the state of the art methods at a significantly faster speed. Besides, our QHNet consumes 50% less mem- ory due to its streamlined architecture. Our code is publicly available as part of the AIRS library (https://github.com/divelab/AIRS).  more » « less
Award ID(s):
2119103
NSF-PAR ID:
10460974
Author(s) / Creator(s):
Date Published:
Journal Name:
International Conference on Machine Learning
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Andreas Krause, Emma Brunskill (Ed.)
    We consider the prediction of the Hamiltonian matrix, which finds use in quantum chemistry and condensed matter physics. Efficiency and equivariance are two important, but conflicting factors. In this work, we propose a SE(3)-equivariant network, named QHNet, that achieves efficiency and equivariance. Our key advance lies at the innovative design of QHNet architecture, which not only obeys the underlying symmetries, but also enables the reduction of number of tensor products by 92%. In addition, QHNet prevents the exponential growth of channel dimension when more atom types are involved. We perform experiments on MD17 datasets, including four molecular systems. Experimental results show that our QHNet can achieve comparable performance to the state of the art methods at a significantly faster speed. Besides, our QHNet consumes 50% less memory due to its streamlined architecture. Our code is publicly available as part of the AIRS library (https://github.com/divelab/AIRS). 
    more » « less
  2. Abstract

    The satellite‐based Cloud Imaging and Particle Size (CIPS) instrument and Atmospheric Infrared Sounder (AIRS) observed concentric gravity waves (GWs) generated by Typhoon Yutu in late October 2018. This work compares CIPS and AIRS nadir viewing observations of GWs at altitudes of 50–55 and 30–40 km, respectively, to simulations from the high‐resolution European Centre for Medium‐Range Weather Forecasting Integrated Forecasting System (ECMWF‐IFS) and ECMWF reanalysis v5 (ERA5). Both ECMWF‐IFS with 9 km and ERA5 with 31 km horizontal resolution show concentric GWs at similar locations and timing as the AIRS and CIPS observations. The GW wavelengths are ∼225–236 km in ECMWF‐IFS simulations, which compares well with the wavelength inferred from the observations. After validation of ECMWF GWs, five category five typhoon events during 2018 are analyzed using ECMWF to obtain characteristics of concentric GWs in the Western Pacific regions. The amplitudes of GWs in the stratosphere are not strongly correlated with the strength of typhoons, but are controlled by background wind conditions. Our results confirm that amplitudes and shapes of concentric GWs observed in the stratosphere and lowermost mesosphere are heavily influenced by the background wind conditions.

     
    more » « less
  3. Supervised machine learning approaches have been increasingly used in accelerating electronic structure prediction as surrogates of first-principle computational methods, such as density functional theory (DFT). While numerous quantum chemistry datasets focus on chemical properties and atomic forces, the ability to achieve accurate and efficient prediction of the Hamiltonian matrix is highly desired, as it is the most important and fundamental physical quantity that determines the quantum states of physical systems and chemical properties. In this work, we generate a new Quantum Hamiltonian dataset, named as QH9, to provide precise Hamiltonian matrices for 2,399 molecular dynamics trajectories and 130,831 stable molecular geometries, based on the QM9 dataset. By designing benchmark tasks with various molecules, we show that current machine learning models have the capacity to predict Hamiltonian matrices for arbitrary molecules. Both the QH9 dataset and the baseline models are provided to the community through an open-source benchmark, which can be highly valuable for developing machine learning methods and accelerating molecular and materials design for scientific and technological applications. Our benchmark is publicly available at \url{https://github.com/divelab/AIRS/tree/main/OpenDFT/QHBench}. 
    more » « less
  4. null (Ed.)
    Abstract. Temperature, H2O, and O3 profiles, as well as CO2, N2O, CH4, chlorofluorocarbon-12 (CFC-12), and sea surface temperature (SST) scalar anomalies are computed using a clear subset of AIRS observations over ocean for the first 16 years of NASA's Earth-Observing Satellite (EOS) Aqua Atmospheric Infrared Sounder (AIRS) operation. The AIRS Level-1c radiances are averaged over 16 d and 40 equal-area zonal bins and then converted to brightness temperature anomalies. Geophysical anomalies are retrieved from the brightness temperature anomalies using a relatively standard optimal estimation approach. The CO2, N2O, CH4, and CFC-12 anomalies are derived by applying a vertically uniform multiplicative shift to each gas in order to obtain an estimate for the gas mixing ratio. The minor-gas anomalies are compared to the National Oceanic and Atmospheric Administration (NOAA) Earth System Research Laboratory (ESRL) in situ values and used to estimate the radiometric stability of the AIRS radiances. Similarly, the retrieved SST anomalies are compared to the SST values used in the ERA-Interim reanalysis and to NOAA's Optimum Interpolation SST (OISST) product. These intercomparisons strongly suggest that many AIRS channels are stable to better than 0.02 to 0.03 K per decade, well below climate trend levels, indicating that the AIRS blackbody is not drifting. However, detailed examination of the anomaly retrieval residuals (observed – computed) shows various small unphysical shifts that correspond to AIRS hardware events (shutdowns, etc.). Some examples are given highlighting how the AIRS radiance stability could be improved, especially for channels sensitive to N2O and CH4. The AIRS shortwave channels exhibit larger drifts that make them unsuitable for climate trending, and they are avoided in this work. The AIRS Level 2 surface temperature retrievals only use shortwave channels. We summarize how these shortwave drifts impacts recently published comparisons of AIRS surface temperature trends to other surface climatologies. 
    more » « less
  5. null (Ed.)
    Abstract Deep neural networks (DNNs) have substantial computational requirements, which greatly limit their performance in resource-constrained environments. Recently, there are increasing efforts on optical neural networks and optical computing based DNNs hardware, which bring significant advantages for deep learning systems in terms of their power efficiency, parallelism and computational speed. Among them, free-space diffractive deep neural networks (D 2 NNs) based on the light diffraction, feature millions of neurons in each layer interconnected with neurons in neighboring layers. However, due to the challenge of implementing reconfigurability, deploying different DNNs algorithms requires re-building and duplicating the physical diffractive systems, which significantly degrades the hardware efficiency in practical application scenarios. Thus, this work proposes a novel hardware-software co-design method that enables first-of-its-like real-time multi-task learning in D 2 2NNs that automatically recognizes which task is being deployed in real-time. Our experimental results demonstrate significant improvements in versatility, hardware efficiency, and also demonstrate and quantify the robustness of proposed multi-task D 2 NN architecture under wide noise ranges of all system components. In addition, we propose a domain-specific regularization algorithm for training the proposed multi-task architecture, which can be used to flexibly adjust the desired performance for each task. 
    more » « less