The Experts below are selected from a list of 285 Experts worldwide ranked by ideXlab platform
Ivars P. Kirsteins - One of the best experts on this subject based on the ideXlab platform.
-
ICASSP - On the structure of the multi-mode filters for passive wavefront curvature ranging in a Distributed Array system
2013 IEEE International Conference on Acoustics Speech and Signal Processing, 2013Co-Authors: Ivars P. KirsteinsAbstract:This work presents some new results on the structures of, and our interpretation on, the multi-rank filters used for passive wavefront curvature (WFC) ranging. Such a WFC ranging systems uses a large-scale Distributed Arrays with many spatially separated modular Arrays, operating under environments subject to a spatial coherence loss. Working on the modular Array level beamformed data, the multi-rank filters along with the weighting coefficients provide further spatial filtering capability to rake in spatial coherence existing in the distorted wavefronts impinging on different modular Arrays. Such multi-rank filters can improve ranging performance through different combining schemes, beyond what achieved by the bearing-only based triangulation. For a real-valued inter-module spatial coherence matrix, the derived multi-rank filters follow a nicely balanced structure comprised of in-phase and quadrature (I/Q) modes with varying spatial directions. The results provide a simple solution for us to discovering levels of coherence existing in different modes for multi-mode combining.
-
ICASSP - Asymptotic results for passive wavefront curvature ranging using a large-scale Distributed Array system
2012 IEEE International Conference on Acoustics Speech and Signal Processing (ICASSP), 2012Co-Authors: Ivars P. KirsteinsAbstract:In this work, we present some asymptotic results on the maximum likelihood multi-rank processor for passive wavefront curvature ranging systems using large-scale Distributed Arrays. We assume that the operation environment for the Distributed Arrays is subject to a spatial coherence loss. Under an exponential coherence model, analytical expressions for the multi-rank combiners are derived. The results provide a simple guideline for choosing inter-module spacing according to the spatial coherence length for formulating multi-rank mode filters and weights used in the combiner. The general framework and the numerical procedures of designing a multi-rank combiner can be applied to other coherence models.
Zhen Xiao - One of the best experts on this subject based on the ideXlab platform.
-
spartan a Distributed Array framework with smart tiling
USENIX Annual Technical Conference, 2015Co-Authors: Chienchin Huang, Qi Chen, Zhaoguo Wang, Russell Power, Jorge Ortiz, Zhen XiaoAbstract:Application programmers in domains like machine learning, scientific computing, and computational biology are accustomed to using powerful, high productivity Array languages such as MatLab, R and NumPy. Distributed Array frameworks aim to scale Array programs across machines. However, maximizing the locality of access to Distributed Arrays is an unsolved problem; such locality is critical for high performance. This paper presents Spartan, a Distributed Array framework that automatically determines how to best partition (aka "tile") n-dimensional Arrays and to co-locate data with computation to maximize locality. Spartan combines a lazy-evaluation based, optimizing frontend with a Distributed tiled Array backend. Central to Spartan's design is a small number of carefully chosen parallel high-level operators, which form the expression graph captured by Spartan's frontend during runtime. These operators simplify the programming of Distributed applications. More importantly, their well-defined semantics allow Spartan's runtime to calculate the costs of different tiling strategies and pick the best one for evaluating the entire expression graph. Using Spartan, we have implemented 12 applications from a variety of domains including machine learning and scientific computing. Our evaluations show that Spartan's automatic tiling mechanism leads to good and scalable performance while eliminating the need for manual tiling.
-
USENIX Annual Technical Conference - Spartan: a Distributed Array framework with smart tiling
2015Co-Authors: Chienchin Huang, Qi Chen, Zhaoguo Wang, Russell Power, Jorge Ortiz, Zhen XiaoAbstract:Application programmers in domains like machine learning, scientific computing, and computational biology are accustomed to using powerful, high productivity Array languages such as MatLab, R and NumPy. Distributed Array frameworks aim to scale Array programs across machines. However, maximizing the locality of access to Distributed Arrays is an unsolved problem; such locality is critical for high performance. This paper presents Spartan, a Distributed Array framework that automatically determines how to best partition (aka "tile") n-dimensional Arrays and to co-locate data with computation to maximize locality. Spartan combines a lazy-evaluation based, optimizing frontend with a Distributed tiled Array backend. Central to Spartan's design is a small number of carefully chosen parallel high-level operators, which form the expression graph captured by Spartan's frontend during runtime. These operators simplify the programming of Distributed applications. More importantly, their well-defined semantics allow Spartan's runtime to calculate the costs of different tiling strategies and pick the best one for evaluating the entire expression graph. Using Spartan, we have implemented 12 applications from a variety of domains including machine learning and scientific computing. Our evaluations show that Spartan's automatic tiling mechanism leads to good and scalable performance while eliminating the need for manual tiling.
Victor Belitsky - One of the best experts on this subject based on the ideXlab platform.
-
Frequency Multiplication in a Distributed Array of SIS Junctions
2014Co-Authors: Bhushan Billade, Alexey Pavolotskiy, Victor BelitskyAbstract:We report the experimental study of the off-chip detection of frequency multiplication in a Distributed Array of Superconductor-Insulator-Superconductor (SIS) junctions. A test device consisting of a series Array consisting sixty eight Nb/Al-AlOx/Nb tunnel junctions was designed for this study, and was fabricated using in-house Nb thin-film technology. The test device with SIS Array was optimized for the study of second harmonic generation in 182−192 GHz output frequency band. The SIS Array was exited with microwave radiation at 3 mm band using a quasi- optically coupled Gunn oscillator and the output response of the device was studied using a double sideband SIS mixer operating in 163 − 211 GHz range with 4−8 GHz. The Josephson-effect for both the SIS multiplier and the detector mixer was carefully suppressed using magnetic field. We observed very sharp second harmonic spectral signals, due to frequency multiplication by the SIS Array. We also observed distinct multi-photon process in the SIS Array tunnel junction response to the applied microwave signal, and the amplitude of the multiplied signal shows dependence on the bias voltage of the SIS Array. We observed that the output power of the multiplied signal increases linearly with the power of the pumping signal up to certain level and them saturates. Increasing the input power beyond this level results in the heating of the chip. When the output of the test device was connected to the LO port of the SIS-mixer, an increase of 10 − 20% in the SIS -mixer dark current was observed when the SIS mixer was voltage biased in the middle of first photon step below the gap voltage. The device, although far from providing sufficient power to pump a practical SIS mixer, may be considered as a first experimental step towards SIS frequency multipliers.
-
Experimental Study of Frequency Multiplication in a Distributed Array of SIS Junctions
IEEE Transactions on Terahertz Science and Technology, 2014Co-Authors: Bhushan Billade, Alexey Pavolotsky, Victor BelitskyAbstract:We report the first experimental off-chip detection of frequency multiplication in a Distributed Array of superconductor-insulator-superconductor (SIS) junctions. A test device consisting of series Array of 68 Nb/Al-AlOx/Nb tunnel junctions was designed to study generation of the second harmonic in the 190-210 GHz band. The SIS Array was exited with microwave radiation at 3-mm band using a quasi-optically coupled Gunn oscillator, and the output response of the device was studied using a double-sideband SIS mixer operating in the 163-211 GHz range with 4-8 GHz IF bandwidth. We measured extremely sharp spectral signals, associated with the ×2 frequency multiplication by the SIS Array. Single- and multi-photon processes were observed in the response of SIS tunnel junction-Array to the applied microwave radiation, confirming device operation in the quantum mode. The output power of the multiplied signal increases linearly with the power of the pumping signal up to certain level and them saturates. In attempt to verify that the device produces noticeable power, the output of the test device was connected to the LO port of the SIS mixer, and an increase of 10%-20% in the SIS mixer dark current was observed. Further development of the demonstrated principle of frequency multiplication may lead to a practical frequency multiplier device.
Chienchin Huang - One of the best experts on this subject based on the ideXlab platform.
-
spartan a Distributed Array framework with smart tiling
USENIX Annual Technical Conference, 2015Co-Authors: Chienchin Huang, Qi Chen, Zhaoguo Wang, Russell Power, Jorge Ortiz, Zhen XiaoAbstract:Application programmers in domains like machine learning, scientific computing, and computational biology are accustomed to using powerful, high productivity Array languages such as MatLab, R and NumPy. Distributed Array frameworks aim to scale Array programs across machines. However, maximizing the locality of access to Distributed Arrays is an unsolved problem; such locality is critical for high performance. This paper presents Spartan, a Distributed Array framework that automatically determines how to best partition (aka "tile") n-dimensional Arrays and to co-locate data with computation to maximize locality. Spartan combines a lazy-evaluation based, optimizing frontend with a Distributed tiled Array backend. Central to Spartan's design is a small number of carefully chosen parallel high-level operators, which form the expression graph captured by Spartan's frontend during runtime. These operators simplify the programming of Distributed applications. More importantly, their well-defined semantics allow Spartan's runtime to calculate the costs of different tiling strategies and pick the best one for evaluating the entire expression graph. Using Spartan, we have implemented 12 applications from a variety of domains including machine learning and scientific computing. Our evaluations show that Spartan's automatic tiling mechanism leads to good and scalable performance while eliminating the need for manual tiling.
-
USENIX Annual Technical Conference - Spartan: a Distributed Array framework with smart tiling
2015Co-Authors: Chienchin Huang, Qi Chen, Zhaoguo Wang, Russell Power, Jorge Ortiz, Zhen XiaoAbstract:Application programmers in domains like machine learning, scientific computing, and computational biology are accustomed to using powerful, high productivity Array languages such as MatLab, R and NumPy. Distributed Array frameworks aim to scale Array programs across machines. However, maximizing the locality of access to Distributed Arrays is an unsolved problem; such locality is critical for high performance. This paper presents Spartan, a Distributed Array framework that automatically determines how to best partition (aka "tile") n-dimensional Arrays and to co-locate data with computation to maximize locality. Spartan combines a lazy-evaluation based, optimizing frontend with a Distributed tiled Array backend. Central to Spartan's design is a small number of carefully chosen parallel high-level operators, which form the expression graph captured by Spartan's frontend during runtime. These operators simplify the programming of Distributed applications. More importantly, their well-defined semantics allow Spartan's runtime to calculate the costs of different tiling strategies and pick the best one for evaluating the entire expression graph. Using Spartan, we have implemented 12 applications from a variety of domains including machine learning and scientific computing. Our evaluations show that Spartan's automatic tiling mechanism leads to good and scalable performance while eliminating the need for manual tiling.
Bhushan Billade - One of the best experts on this subject based on the ideXlab platform.
-
Frequency Multiplication in a Distributed Array of SIS Junctions
2014Co-Authors: Bhushan Billade, Alexey Pavolotskiy, Victor BelitskyAbstract:We report the experimental study of the off-chip detection of frequency multiplication in a Distributed Array of Superconductor-Insulator-Superconductor (SIS) junctions. A test device consisting of a series Array consisting sixty eight Nb/Al-AlOx/Nb tunnel junctions was designed for this study, and was fabricated using in-house Nb thin-film technology. The test device with SIS Array was optimized for the study of second harmonic generation in 182−192 GHz output frequency band. The SIS Array was exited with microwave radiation at 3 mm band using a quasi- optically coupled Gunn oscillator and the output response of the device was studied using a double sideband SIS mixer operating in 163 − 211 GHz range with 4−8 GHz. The Josephson-effect for both the SIS multiplier and the detector mixer was carefully suppressed using magnetic field. We observed very sharp second harmonic spectral signals, due to frequency multiplication by the SIS Array. We also observed distinct multi-photon process in the SIS Array tunnel junction response to the applied microwave signal, and the amplitude of the multiplied signal shows dependence on the bias voltage of the SIS Array. We observed that the output power of the multiplied signal increases linearly with the power of the pumping signal up to certain level and them saturates. Increasing the input power beyond this level results in the heating of the chip. When the output of the test device was connected to the LO port of the SIS-mixer, an increase of 10 − 20% in the SIS -mixer dark current was observed when the SIS mixer was voltage biased in the middle of first photon step below the gap voltage. The device, although far from providing sufficient power to pump a practical SIS mixer, may be considered as a first experimental step towards SIS frequency multipliers.
-
Experimental Study of Frequency Multiplication in a Distributed Array of SIS Junctions
IEEE Transactions on Terahertz Science and Technology, 2014Co-Authors: Bhushan Billade, Alexey Pavolotsky, Victor BelitskyAbstract:We report the first experimental off-chip detection of frequency multiplication in a Distributed Array of superconductor-insulator-superconductor (SIS) junctions. A test device consisting of series Array of 68 Nb/Al-AlOx/Nb tunnel junctions was designed to study generation of the second harmonic in the 190-210 GHz band. The SIS Array was exited with microwave radiation at 3-mm band using a quasi-optically coupled Gunn oscillator, and the output response of the device was studied using a double-sideband SIS mixer operating in the 163-211 GHz range with 4-8 GHz IF bandwidth. We measured extremely sharp spectral signals, associated with the ×2 frequency multiplication by the SIS Array. Single- and multi-photon processes were observed in the response of SIS tunnel junction-Array to the applied microwave radiation, confirming device operation in the quantum mode. The output power of the multiplied signal increases linearly with the power of the pumping signal up to certain level and them saturates. In attempt to verify that the device produces noticeable power, the output of the test device was connected to the LO port of the SIS mixer, and an increase of 10%-20% in the SIS mixer dark current was observed. Further development of the demonstrated principle of frequency multiplication may lead to a practical frequency multiplier device.