-
公开(公告)号:US10341079B2
公开(公告)日:2019-07-02
申请号:US15834739
申请日:2017-12-07
Applicant: Microsoft Technology Licensing, LLC
Inventor: Chen Zhang
Abstract: The described technology provides front-end antenna architecture for wireless communication that manages multiple frequency sub-bands in a manner that results in a low insertion loss rate. Further, the components of the architecture are smaller than typical wireless communication antenna architectures, and therefore the described architecture can be accommodated in a smaller area than typical architectures without a decrease in efficiency. A matching circuit is communicatively connected to each frequency sub-band's respective signal path. When two or more sub-bands are requested for high-speed wireless communication, the matching circuits match the impedance of each sub-band with one-another. Matching the impedance allows two or more sub-bands to be used to wirelessly communicate while maintaining high efficiency. The matching circuits are disabled when only one sub-band is needed for wireless communication.
-
公开(公告)号:US09871646B2
公开(公告)日:2018-01-16
申请号:US14871219
申请日:2015-09-30
Applicant: Microsoft Technology Licensing, LLC
Inventor: Chen Zhang
Abstract: The described technology provides front-end antenna architecture for wireless communication that manages multiple frequency sub-bands in a manner that results in a low insertion loss rate. Further, the components of the architecture are smaller than typical wireless communication antenna architectures, and therefore the described architecture can be accommodated in a smaller area than typical architectures without a decrease in efficiency. A matching circuit is communicatively connected to each frequency sub-band's respective signal path. When two or more sub-bands are requested for high-speed wireless communication, the matching circuits match the impedance of each sub-band with one another. Matching the impedance allows two or more sub-bands to be used to wirelessly communicate while maintaining high efficiency. The matching circuits are disabled when only one sub-band is needed for wireless communication.
-
公开(公告)号:US11763156B2
公开(公告)日:2023-09-19
申请号:US16686089
申请日:2019-11-15
Applicant: Microsoft Technology Licensing, LLC
Inventor: Chen Zhang , Yunxin Liu
CPC classification number: G06N3/082
Abstract: In embodiments of the present disclosure, there is provided an approach for neural network model compression based on bank-balanced sparsity. In embodiments of the present disclosure, a set of weight parameters, such as a weight matrix, in a neural network is divided into a plurality of equal-sized banks in terms of number of elements, and then all of the equal-sized banks are pruned at the same sparsity level. In this way, each pruned bank will have the same number of non-zero elements, which is suitable for hardware speedup. Moreover, since each bank is pruned independently in a fine granularity, the model accuracy can be ensured. Thus, according to embodiments of the present disclosure, the neural network compression method based on bank-balanced sparsity can achieve both high model accuracy and high hardware speedup.
-
公开(公告)号:US10320131B2
公开(公告)日:2019-06-11
申请号:US15628427
申请日:2017-06-20
Applicant: Microsoft Technology Licensing, LLC
Inventor: Chen Zhang , Toby Morris
IPC: H01R29/00 , H01R24/50 , H01R13/703 , H01R13/74 , H01R24/44
Abstract: Circuit boards incorporating one or more wireless communication components are often subjected to certification testing during commissioning. A test port is often included within each circuit board to perform the certification testing. The presently disclosed recessed test connectors reduce or eliminate protrusion of the test port on a non-component side of the circuit board, while still providing access to the test port from the non-component side of the circuit board.
-
公开(公告)号:US20220245433A1
公开(公告)日:2022-08-04
申请号:US17615077
申请日:2020-04-29
Applicant: Microsoft Technology Licensing, LLC
Inventor: Chen Zhang , Yunxin LIU
Abstract: Various implementations of the subject matter as described herein relate to a sparse convolutional neural network. In some implementations, a computer-implemented method comprises: quantizing an input feature map to obtain a quantized input feature map; determining, based on the quantized input feature map, a sparsity mask for an output feature map through a quantized version of a convolutional neural network, the sparsity mask indicating positions of non-zero entries in the output feature map; and determining, based on the input feature map, the non-zero entries indicated by the sparsity mask in the output feature map through the convolutional neural network.
-
公开(公告)号:US20180123770A1
公开(公告)日:2018-05-03
申请号:US15834739
申请日:2017-12-07
Applicant: Microsoft Technology Licensing, LLC
Inventor: Chen Zhang
Abstract: The described technology provides front-end antenna architecture for wireless communication that manages multiple frequency sub-bands in a manner that results is a low insertion loss rate. Further, the components of the architecture are smaller than typical wireless communication antenna architectures, and therefore the described architecture can be accommodated in a smaller area than typical architectures without a decrease in efficiency. A matching circuit is communicatively connected to each frequency sub-band's respective signal path. When two or more sub-bands are requested for high-speed wireless communication, the matching circuits match the impedance of each sub-band with one-another. Matching the impedance allows two or more sub-bands to be used to wirelessly communicate while maintaining high efficiency. The matching circuits are disabled when only one sub-band is needed for wireless communication.
-
公开(公告)号:US20170207534A1
公开(公告)日:2017-07-20
申请号:US15001008
申请日:2016-01-19
Applicant: Microsoft Technology Licensing, LLC
Inventor: Chen Zhang
IPC: H01Q5/335
Abstract: The described technology provides an antenna interface for wireless communication that allows the use of a single transmission line trace between at least one antenna and a transceiver of a wireless communications device. The components of the antenna interface include a first set of filters, a set of low noise amplifiers (LNAs) and a second set of filters. The antenna interface, which is communicatively coupled to the single transmission line trace, avoids the use of a coaxial cable and reduces issues related to the use of multiple transmission line traces.
-
公开(公告)号:US20170093552A1
公开(公告)日:2017-03-30
申请号:US14871219
申请日:2015-09-30
Applicant: Microsoft Technology Licensing, LLC
Inventor: Chen Zhang
Abstract: The described technology provides front-end antenna architecture for wireless communication that manages multiple frequency sub-bands in a manner that results is a low insertion loss rate. Further, the components of the architecture are smaller than typical wireless communication antenna architectures, and therefore the described architecture can be accommodated in a smaller area than typical architectures without a decrease in efficiency. A matching circuit is communicatively connected to each frequency sub-band's respective signal path. When two or more sub-bands are requested for high-speed wireless communication, the matching circuits match the impedance of each sub-band with one-another. Matching the impedance allows two or more sub-bands to be used to wirelessly communicate while maintaining high efficiency. The matching circuits are disabled when only one sub-band is needed for wireless communication.
-
公开(公告)号:US20210150362A1
公开(公告)日:2021-05-20
申请号:US16686089
申请日:2019-11-15
Applicant: Microsoft Technology Licensing, LLC
Inventor: Chen Zhang , Yunxin Liu
Abstract: In embodiments of the present disclosure, there is provided an approach for neural network model compression based on bank-balanced sparsity. In embodiments of the present disclosure, a set of weight parameters, such as a weight matrix, in a neural network is divided into a plurality of equal-sized banks in terms of number of elements, and then all of the equal-sized banks are pruned at the same sparsity level. In this way, each pruned bank will have the same number of non-zero elements, which is suitable for hardware speedup. Moreover, since each bank is pruned independently in a fine granularity, the model accuracy can be ensured. Thus, according to embodiments of the present disclosure, the neural network compression method based on bank-balanced sparsity can achieve both high model accuracy and high hardware speedup.
-
-
-
-
-
-
-
-