A5100765709- Parallel Computing and Optimization Techniques
- Cloud Computing and Resource Management
- Advanced Optical Network Technologies
- Geoscience and Mining Technology
- Advanced Neural Network Applications
- Geomechanics and Mining Engineering
- Hydrocarbon exploration and reservoir analysis
- Software-Defined Networks and 5G
- VLSI and Analog Circuit Testing
- Tunneling and Rock Mechanics
- Geological and Geochemical Analysis
- Advanced Measurement and Detection Methods
- Software Engineering Techniques and Practices
- Stochastic Gradient Optimization Techniques
- Coal Properties and Utilization
- Optical Network Technologies
- Interconnection Networks and Systems
- Advanced Data Storage Technologies
- Rock Mechanics and Modeling
- Enhanced Oil Recovery Techniques
- Engineering and Test Systems
- Hydraulic Fracturing and Reservoir Analysis
- Structural Engineering and Vibration Analysis
- Civil and Geotechnical Engineering Research
- Photonic and Optical Devices
Google (United States)
2020-2024
Jingdong (China)
2023
Liaoning Technical University
2020
Jiangxi University of Science and Technology
2020
Saskatchewan Research Council (Canada)
2017
Intel (United States)
2016
Google's TPU supercomputers train deep neural networks 50x faster than general-purpose running a high-performance computing benchmark.
In response to innovations in machine learning (ML) models, production workloads changed radically and rapidly. TPU v4 is the fifth Google domain specific architecture (DSA) its third supercomputer for such ML models. Optical circuit switches (OCSes) dynamically reconfigure interconnect topology improve scale, availability, utilization, modularity, deployment, security, power, performance; users can pick a twisted 3D torus if desired. Much cheaper, lower faster than Infiniband, OCSes...
Five years ago, few would have predicted that a software company like Google build its own computers. Nevertheless, has been deploying computers for machine learning (ML) training since 2017, powering key services. These Tensor Processing Units (TPUs) are composed of chips, systems, and software, all co-designed in-house. In this paper, we detail the circumstances led to outcome, challenges opportunities observed, approach taken quick review performance, finally retrospective on results. A...
We describe our experience developing what we believe to be the world's first large-scale production deployments of lightwave fabrics used for both datacenter networking and machine-learning (ML) applications. Using optical circuit switches (OCSes) transceivers developed in-house, employ hardware software codesign integrate into network computing infrastructure. Key design is a high degree multiplexing enabled by new kinds wavelength-division-multiplexing (WDM) circulators that support...
Abstract The association between mineralisation and hydrocarbons in sedimentary basins is widely recognised, but the nature significance of their relationships are not fully understood. This paper provides an example metalliferous brines injecting into a palaeo‐oil–gas reservoir to form world‐class Zn–Pb deposit (Jinding, China). Petrographic microthermometric studies oil inclusions PVT simulations suggest that gas were charged shallow (<1300 m) environment before mineralisation. favoured...
The permeability enhancement is the key to increase efficiency of gas drainage from low-permeability coal seam. In this paper, acid fracturing technology proposed solve problem low seam with high mineral content. self-developed coal-rock experimental system used study whole process under different stress conditions and specimens varying enhancing methods. surface characteristics content before after were tested by SEM XRF technologies. test results show that effect acidizing mainly dominated...
Distributed in-memory key-value stores (KVSs) have become a critical data-serving layer in cloud computing and big data infrastructure. Unfortunately, KVSs demonstrated gap between achieved available performance, QoS, energy efficiency on commodity platforms. Two research thrusts focused improving performance: hardware-centric has started to explore specialized platforms for KVSs, software-centric revisited the KVS application address fundamental software bottlenecks. Unlike prior focusing...
We describe our experience developing what we believe to be the world's first large-scale production deployments of lightwave fabrics used for both datacenter networking and machine-learning (ML) applications [1],[2],[3],[4]. Using optical circuit switches (OCSes) transceivers developed in-house, employ hardware software codesign integrate into network computing infrastructure. Key design is a high degree multiplexing enabled by new kinds wavelength-division-multiplexing (WDM) circulators...
<title>Abstract</title> The structure of a large museum is novel, with zigzag-shaped steel suspension bridge system in the atrium. To study impact this on overall seismic performance museum, ABAQUS software was used to create two 3D models: Model I without and II system. An in-depth analysis conducted time-history response laws maximum roof displacement, inter-story drift angle, acceleration under frequently rarely occurring earthquake conditions, providing comprehensive performance. found...
In this paper, through the statistical analysis of data contact resistance changing with time when circuit intermittent fault occurs, a model is proposed, and an injector built by FPGA high-speed programmable resistor, which use to inject into circuit, so as carry out research related fault. Then, application illustrated example health status recognition. example, used keep circuits in different states, then FSV method extract feature fault, after training HSMM models, state identified....
When the T/R-R radar imaging system is used for fusion of space targets, effective synthetic aperture size closely related to station deployment. In order improve efficiency system, a method optimizing deployment receiving R studied. This fully utilizes prior information three-axis stability targets in space, studies determining monostatic/bistatic ISAR plane, and then determines plane when T/R stations observe alone, as well bistatic together; Then, taking angle between two planes objective...