检索结果-内蒙古大学图书馆

Deep Learning and Spectral Embedding for Graph Partitioning

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Gatti, Alice Hu, Zhixiong Smidt, Tess Ng, Esmond G. Ghysels, Pieter Computational Research Division Lawrence Berkeley National Laboratory BerkeleyCA United States University of California Santa CruzCA United States MIT Department of Electrical Engineering and Computer Science CambridgeMA United States

We present a graph bisection and partitioning algorithm based on graph neural networks. For each node in the graph, the network outputs probabilities for each of the partitions. The graph neural network consists of two modules: an embedding phase and a partitioning phase. The embedding phase is trained first by minimizing a loss function inspired by spectral graph theory. The partitioning module is trained through a loss function that corresponds to the expected value of the normalized cut. Both parts of the neural network rely on SAGE convolutional layers and graph coarsening using heavy edge matching. The multilevel structure of the neural network is inspired by the multigrid algorithm. Our approach generalizes very well to bigger graphs and has partition quality comparable to METIS, Scotch and spectral partitioning, with shorter runtime compared to METIS and spectral partitioning. Copyright © 2021, The Authors. All rights reserved.

关键词： Graph neural networks

Realization of High-Fidelity CZ and ZZ-Free iSWAP Gates with a Tunable Coupler

学校读者我要写书评

暂无评论

Physical Review X 2021年第2期11卷 021058-021058页

作者： Youngkyu Sung Leon Ding Jochen Braumüller Antti Vepsäläinen Bharath Kannan Morten Kjaergaard Ami Greene Gabriel O. Samach Chris McNally David Kim Alexander Melville Bethany M. Niedzielski Mollie E. Schwartz Jonilyn L. Yoder Terry P. Orlando Simon Gustavsson William D. Oliver Research Laboratory of Electronics Massachusetts Institute of Technology Cambridge Massachusetts 02139 USA Department of Electrical Engineering and Computer Science Massachusetts Institute of Technology Cambridge Massachusetts 02139 USA Department of Physics Massachusetts Institute of Technology Cambridge Massachusetts 02139 USA MIT Lincoln Laboratory Lexington Massachusetts 02421 USA

High-fidelity two-qubit gates at scale are a key requirement to realize the full promise of quantum computation and simulation. The advent and use of coupler elements to tunably control two-qubit interactions has improved operational fidelity in many-qubit systems by reducing parasitic coupling and frequency crowding issues. Nonetheless, two-qubit gate errors still limit the capability of near-term quantum applications. The reason, in part, is that the existing framework for tunable couplers based on the dispersive approximation does not fully incorporate three-body multilevel dynamics, which is essential for addressing coherent leakage to the coupler and parasitic longitudinal (ZZ) interactions during two-qubit gates. Here, we present a systematic approach that goes beyond the dispersive approximation to exploit the engineered level structure of the coupler and optimize its control. Using this approach, we experimentally demonstrate CZ and ZZ-free iSWAP gates with two-qubit interaction fidelities of 99.76±0.07% and 99.87±0.23%, respectively, which are close to their T1 limits.

关键词： Cavity quantum electrodynamics Quantum control Quantum gates Quantum information with solid state qubits Quantum optics with artificial atoms Superconducting quantum optics

Physion: Evaluating Physical Prediction from Vision in Humans and Machines 35

学校读者我要写书评

暂无评论

Physion: Evaluating Physical Prediction from Vision in Human...

35th Conference on Neural Information Processing Systems - Track on Datasets and Benchmarks, NeurIPS Datasets and Benchmarks 2021

作者： Bear, Daniel M. Wang, Elias Mrowca, Damian Binder, Felix Tung, Hsiao-Yu Fish Pramod, R.T. Holdaway, Cameron Tao, Sirui Smith, Kevin Sun, Fan-Yun Fei-Fei, Li Kanwisher, Nancy Tenenbaum, Joshua B. Yamins, Daniel L.K. Fan, Judith Department of Psychology StanfordCA94305 United States Department of Electrical Engineering StanfordCA94305 United States Department of Computer Science StanfordCA94305 United States Wu Tsai Neurosciences Institute StanfordCA94305 United States Department of Cognitive Science UC San DiegoCA92093 United States Department of Psychology UC San DiegoCA92093 United States Department of Brain and Cognitive Sciences CBMM MIT CambridgeMA02139 United States

While current vision algorithms excel at many challenging tasks, it is unclear how well they understand the physical dynamics of real-world environments. Here we introduce Physion, a dataset and benchmark for rigorously evaluating the ability to predict how physical scenarios will evolve over time. Our dataset features realistic simulations of a wide range of physical phenomena, including rigid and soft-body collisions, stable multi-object configurations, rolling, sliding, and projectile motion, thus providing a more comprehensive challenge than previous benchmarks. We used Physion to benchmark a suite of models varying in their architecture, learning objective, input-output structure, and training data. In parallel, we obtained precise measurements of human prediction behavior on the same set of scenarios, allowing us to directly evaluate how well any model could approximate human behavior. We found that vision algorithms that learn object-centric representations generally outperform those that do not, yet still fall far short of human performance. On the other hand, graph neural networks with direct access to physical state information both perform substantially better and make predictions that are more similar to those made by humans. These results suggest that extracting physical representations of scenes is the main bottleneck to achieving human-level and human-like physical understanding in vision algorithms. We have publicly released all data and code to facilitate the use of Physion to benchmark additional models in a fully reproducible manner, enabling systematic evaluation of progress towards vision algorithms that understand physical environments as robustly as people do. © 2021 Neural information processing systems foundation. All rights reserved.

关键词： Graph neural networks

Pale Red Dot: A Large, Robust Architecture for Human Settlements on Mars

学校读者我要写书评

暂无评论

Pale Red Dot: A Large, Robust Architecture for Human Settlem...

Accelerating Space Commerce, Exploration, and New Discovery Conference, ASCEND 2023

作者： Lordos, George Hoying, Madelyn McKinney, Lanie Webb, Alisa Lin, Yuying Martin, Estelle Beilstein, John Miller, Duncan Villegas, David Sepulveda, Francisco AlSadah, Yousif Iyer, Divya O’Neill, Cormac Patel, Palak Garcia, Ignacio Arzuaga Azzouz, H. Chan, Wing Lam Eyre, Ezra Gleason, Dane Romero, Elizabeth Arias, Liliana Orozco, Omar Gentgen, Chloe Latyshev, Kir Hoffman, Jeffrey de Weck, Olivier Ikinci, Meltem Lordos, Alexandros Aeronautics and Astronautics Massachusetts Institute of Technology 33-409 CambridgeMA02139 United States Massachusetts Institute of Technology Harvard-MIT Health Sciences and Technology Program and MIT Aeronautics and Astronautics 33-409 CambridgeMA02139 United States Massachusetts Institute of Technology CambridgeMA02139 United States Aeronautics and Astronautics Massachusetts Institute of Technology 77 Mass Ave. 33-409 CambridgeMA02139 United States System Design and Management Massachusetts Institute of Technology 77 Mass Ave. E40-300 CambridgeMA02139 United States Mechanical Engineering Massachusetts Institute of Technology 77 Mass Ave. 33-409 CambridgeMA02139 United States Civil and Environmental Engineering Massachusetts Institute of Technology 77 Mass Ave. CambridgeMA02139 United States Electrical Engineering Massachusetts Institute of Technology 77 Mass Ave. CambridgeMA02139 United States Nuclear Science and Engineering Massachusetts Institute of Technology 77 Mass Ave. CambridgeMA02139 United States Computer Science and Engineering Massachusetts Institute of Technology 77 Mass Ave. CambridgeMA02139 United States Massachusetts Institute of Technology 77 Mass Ave. 37-219 CambridgeMA02139 United States Massachusetts Institute of Technology 77 Mass Ave. 33-410 CambridgeMA02139 United States Behavioral Economics University of Cyprus Nicosia Cyprus Department of Psychology University of Cyprus Nicosia Cyprus

ISBN: (数字)9781624107054

ISBN: (纸本)9781624107054

We propose a ten-yearMars mission to establish Earth-independence and support efficient, safe, and sustainable science and exploration goals. The campaign begins with a series of precursor missions, starting with the Artemis program, followed by robotic missions to performsite surveys and selection, de-risk technologies, and to start stockpiling in-situ water and propellant. Robotic Cargo Starships would depart first in 2035 to deliver and deploy approximately 5,800 tons of habitats, nuclear microreactors, farmmodules, manufacturing and ISRU systems. In 2040, two Crew Starships carrying 36 persons would land onMars to establish two close-proximity villages. The lifecycle cost is estimated at $81 billion, with a peak annual cost of $6.6 billion. To meet RASC-AL competition requirements, intended to motivate the exploration of highly self-reliant architecture concepts, resupply from Earth was constrained to only 5 metric tons every two years and minimum surface endurance was set at 7 years. The risks to human life and to critical systems increase exponentially over a minimally-resupplied, 7-10 year mission toMars. Our approach focused on identifying capabilities to mitigate known and unknown risks. Using a high-fidelity crew time model, we found that large architectures, with their economies of scale and productivity gains from specialization, are essential to robustly support the necessary capabilities. The model also shows that small 7-year missions of 12 persons or less cannot provide value in terms of free surface traverse time for science and exploration. Given the necessity of a large, complex architecture, we then sought out efficiencies and synergies (value engineering) in order to make the mission feasible and affordable. In the process, we identified more benefits that are enabled only by large missions, including the creation of a true community on Mars with sufficient social complexity for humans to thrive, and the geopolitically compelling option to inclu

关键词： Value engineering

pyscreener: A Python Wrapper for Computational Docking Software

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Graff, David E. Coley, Connor W. Department of Chemistry and Chemical Biology Harvard University CambridgeMA United States Department of Chemical Engineering Massachusetts Institute of Technology CambridgeMA United States Department of Electrical Engineering and Computer Science Massachusetts Institute of Technology MIT CambridgeMA United States

pyscreener is a Python library that seeks to alleviate the challenges of large-scale structure-based design using computational docking. It provides a simple and uniform interface that is agnostic to the backend docking engine with which to calculate the docking score of a given molecule in a specified active site. Additionally, pyscreener features first-class support for task distribution, allowing users to seamlessly scale their code from a local, multi-core setup to a large, heterogeneous resource allocation. Copyright © 2021, The Authors. All rights reserved.

关键词： Python

Probabilistic Genotype-Phenotype Maps Reveal Mutational Robustness of RNA Folding, Spin Glasses, and Quantum Circuits

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Sappington, Anna Mohanty, Vaibhav Department of Electrical Engineering and Computer Science Massachusetts Institute of Technology CambridgeMA02139 United States Program for Health Sciences and Technology Harvard Medical School BostonMA02115 United States Massachusetts Institute of Technology CambridgeMA02139 United States Harvard/MIT MD-PhD Program Harvard Medical School BostonMA02115 United States Department of Chemistry and Chemical Biology Harvard University CambridgeMA02138 United States

Recent studies of genotype-phenotype (GP) maps have reported universally enhanced phenotypic robustness to genotype mutations, a feature essential to evolution. Virtually all of these studies make a simplifying assumption that each genotype—represented as a sequence—maps deterministically to a single phenotype, such as a discrete structure. Here, we introduce probabilistic genotype-phenotype (PrGP) maps, where each genotype maps to a vector of phenotype probabilities, as a more realistic and universal language for investigating robustness in a variety of physical, biological, and computational systems. We study three model systems to show that PrGP maps offer a generalized framework which can handle uncertainty emerging from various physical sources: (1) thermal fluctuation in RNA folding, (2) external field disorder in spin glass ground state finding, and (3) superposition and entanglement in quantum circuits, which are realized experimentally on IBM quantum computers. In all three cases, we observe a novel biphasic robustness scaling which is enhanced relative to random expectation for more frequent phenotypes and approaches random expectation for less frequent phenotypes. We derive an analytical theory for the behavior of PrGP robustness, and we demonstrate that the theory is highly predictive of empirical robustness. © 2023, CC BY.

关键词： Quantum computers

Evolution of 1/f Flux Noise in Superconducting Qubits with Weak Magnetic Fields

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Rower, David A. Ateshian, Lamia Li, Lauren H. Hays, Max Bluvstein, Dolev Ding, Leon Kannan, Bharath Almanakly, Aziza Braumüller, Jochen Kim, David K. Melville, Alexander Niedzielski, Bethany M. Schwartz, Mollie E. Yoder, Jonilyn L. Orlando, Terry P. Wang, Joel I-Jan Gustavsson, Simon Grover, Jeffrey A. Serniak, Kyle Comin, Riccardo Oliver, William D. Department of Physics Massachusetts Institute of Technology CambridgeMA02139 United States Research Laboratory of Electronics Massachusetts Institute of Technology CambridgeMA02139 United States Department of Electrical Engineering and Computer Science Massachusetts Institute of Technology CambridgeMA02139 United States Department of Physics Harvard University CambridgeMA02139 United States MIT Lincoln Laboratory LexingtonMA02421 United States Atlantic Quantum CambridgeMA02139 United States

The microscopic origin of 1/f magnetic flux noise in superconducting circuits has remained an open question for several decades despite extensive experimental and theoretical investigation. Recent progress in superconducting devices for quantum information has highlighted the need to mitigate sources of qubit decoherence, driving a renewed interest in understanding the underlying noise mechanism(s). Though a consensus has emerged attributing flux noise to surface spins, their identity and interaction mechanisms remain unclear, prompting further study. Here we apply weak in-plane magnetic fields to a capacitively-shunted flux qubit (where the Zeeman splitting of surface spins lies below the device temperature) and study the flux-noise-limited qubit dephasing, revealing previously unexplored trends that may shed light on the dynamics behind the emergent 1/f noise. Notably, we observe an enhancement (suppression) of the spin-echo (Ramsey) pure dephasing time in fields up to B = 100 G. With direct noise spectroscopy, we further observe a transition from a 1/f to approximately Lorentzian frequency dependence below 10 Hz and a reduction of the noise above 1 MHz with increasing magnetic field. We suggest that these trends are qualitatively consistent with an increase of spin cluster sizes with magnetic field. These results should help to inform a complete microscopic theory of 1/f flux noise in superconducting circuits. Copyright © 2023, The Authors. All rights reserved.

关键词： Magnetic fields

AI-driven 3D Spatial Transcriptomics

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Almagro-Pérez, Cristina Song, Andrew H. Weishaupt, Luca Kim, Ahrong Jaume, Guillaume Williamson, Drew F.K. Hemker, Konstantin Lu, Ming Y. Singh, Kritika Chen, Bowen Le, Long Phi Baras, Alexander S. Jiang, Sizun Bashashati, Ali Liu, Jonathan T.C. Mahmood, Faisal Department of Pathology Mass General Brigham Harvard Medical School BostonMA United States Cancer Program Broad Institute of Harvard and MIT CambridgeMA United States Data Science Program Dana-Farber Cancer Institute BostonMA United States Harvard-MIT Division of Health Sciences and Technology Massachusetts Institute of Technology CambridgeMA United States Department of Pathology Pusan National University Busan Korea Republic of Department of Pathology and Laboratory Medicine Emory University School of Medicine AtlantaGA United States Department of Computer Science & Technology University of Cambridge Cambridge United Kingdom Department of Electrical Engineering and Computer Science Massachusetts Institute of Technology CambridgeMA United States Rutgers Robert Wood Johnson Medical School New BrunswickNJ United States Department of Pathology Johns Hopkins Hospital BaltimoreMD United States Department of Biomedical Engineering Johns Hopkins Hospital BaltimoreMD United States Broad Institute of Harvard and MIT CambridgeMA United States Center for Virology and Vaccine Research Beth Israel Deaconess Medical Center Harvard Medical School BostonMA United States Department of Pathology Dana Farber Cancer Institute BostonMA United States Department of Pathology and Laboratory Medicine University of British Columbia VancouverBC Canada School of Biomedical Engineering University of British Columbia VancouverBC Canada Department of Mechanical Engineering Bioengineering Laboratory Medicine & Pathology University of Washington SeattleWA United States Harvard Data Science Initiative Harvard University CambridgeMA United States

A comprehensive three-dimensional (3D) map of tissue architecture and gene expression is crucial for illuminating the complexity and heterogeneity of tissues across diverse biomedical applications1. However, most spatial transcriptomics (ST) approaches remain limited to two-dimensional (2D) sections of tissue2–4. Although current 3D ST methods hold promise, they typically require extensive tissue sectioning, are complex, are not compatible with non-destructive 3D tissue imaging technologies, and often lack scalability5–8. Here, we present VOlumetrically Resolved Transcriptomics EXpression (VORTEX), an AI framework that leverages 3D tissue morphology and minimal 2D ST to predict volumetric 3D ST. By pretraining on diverse 3D morphology–transcriptomic pairs from heterogeneous tissue samples and then fine-tuning on minimal 2D ST data from a specific volume of interest, VORTEX learns both generic tissue-related and sample-specific morphological correlates of gene expression. This approach enables dense, high-throughput, and fast 3D ST, scaling seamlessly to large tissue volumes far beyond the reach of existing 3D ST techniques. By offering a cost-effective and minimally destructive route to obtaining volumetric molecular insights, we anticipate that VORTEX will accelerate biomarker discovery and our understanding of morphomolecular associations and cell states in complex tissues. Interactive 3D ST volumes can be viewed at https://***/. © 2025, CC BY-NC-ND.

关键词： Vortex flow

Quantifying disparities in intimate partner violence: a machine learning method to correct for underreporting

学校读者我要写书评

暂无评论

npj Women's Health

npj Women's Health 2024年第1期2卷 1-13页

作者： Divya Shanmugam Kaihua Hou Emma Pierson Department of Electrical Engineering and Computer Science MIT Cambridge MA 02139 USA Malone Center of Engineering in Healthcare Johns Hopkins University School of Medicine Baltimore MD USA Department of Computer Science Cornell Tech New York NY 10044 USA Department of Population Health Sciences Weill Cornell Medical College New York NY 10021 USA

The first step towards reducing the pervasive disparities in women’s health is to quantify them. Accurate estimates of therelative prevalenceacross groups—capturing, for example, that a condition affects Black women more frequently than white women—facilitate effective and equitable health policy that prioritizes groups who are disproportionately affected by a condition. However, it is difficult to estimate relative prevalence when a health condition is underreported, as many women’s health conditions are. In this work, we presentPURPLE, a method for accurately estimating the relative prevalence of underreported health conditions which builds upon the literature in positive unlabeled learning. We show that under a commonly made assumption—that the probability of having a health condition given a set of symptoms remains constant across groups—we can recover the relative prevalence, even without restrictive assumptions commonly made in positive unlabeled learning and even if it is impossible to recover the absolute prevalence. We conduct experiments on synthetic and real health data which demonstratePURPLE’s ability to recover the relative prevalence more accurately than do previous methods. We then usePURPLEto quantify the relative prevalence of intimate partner violence (IPV) in two large emergency department datasets. We find higher prevalences of IPV among patients who are on Medicaid, not legally married, and non-white, and among patients who live in lower-income zip codes or in metropolitan counties. We show that correcting for underreporting is important to accurately quantify these disparities and that failing to do so yields less plausible estimates. Our method is broadly applicable to underreported conditions in women’s health, as well as to gender biases beyond healthcare.

关键词：