专利快速检索-快速检索全球专利，免费商用专利数据库-IPRDB

1. 发明申请

WO2020046607A1 COMPUTING DEVICE FOR MULTIPLE ACTIVATION FUNCTIONS IN NEURAL NETWORKS 审中-公开
公开(公告)号：WO2020046607A1
公开(公告)日：2020-03-05
申请号：PCT/US2019/046998
申请日：2019-08-19
申请人： DINOPLUSAI HOLDINGS LIMITED , CHIN, Chung Kuang , WU, Tong , SABER, Ahmed , SERTILLANGE, Steven
发明人： CHIN, Chung Kuang , WU, Tong , SABER, Ahmed , SERTILLANGE, Steven
IPC分类号： G06N3/02 , G06N3/04 , G06N3/06 , G06N3/063 , G06N3/067 , G06N3/08 , G06N3/10 , G06N3/12
摘要： A scalar element computing device for computing a selected activation function selected from two or more different activation functions is disclosed. The scalar element computing device comprises N processing elements, N command memories and an operator pool. The N processing elements are arranged into a pipeline to cause the outputs of each non-last-stage processing element coupled to the inputs of one next-stage processing element. The N command memories are coupled to the N processing elements individually. The operator pool is coupled to the N processing elements, where the operator pool comprises a set of operators for implementing any activation function in an activation function group. The N processing elements are configured according to command information stored in the N command memories to calculate a target activation function selected from the activation function group by using one or more operators in the set of operations.

2. 发明申请

WO2019143537A1 MISSION-CRITICAL AI PROCESSOR WITH RECORD AND REPLAY SUPPORT 审中-公开
公开(公告)号：WO2019143537A1
公开(公告)日：2019-07-25
申请号：PCT/US2019/013404
申请日：2019-01-13
申请人： DINOPLUSAI HOLDINGS LIMITED , HU, Yujie , WU, Tong , WANG, Xiaosong , ZHU, Zongwei , CHIN, Chung Kuang , GOLD, Clifford , SERTILLANGE, Steven , WONG, Yick Kei
发明人： HU, Yujie , WU, Tong , WANG, Xiaosong , ZHU, Zongwei , CHIN, Chung Kuang , GOLD, Clifford , SERTILLANGE, Steven , WONG, Yick Kei
IPC分类号： G06F9/38 , G06F13/28 , G06F15/18 , G06N3/02 , G06N99/00
摘要： Embodiments described herein provide a mission-critical artificial intelligence (AI) processor (MAIP), which includes an instruction buffer, processing circuitry, a data buffer, command circuitry, and communication circuitry. During operation, the instruction buffer stores a first hardware instruction and a second hardware instruction. The processing circuitry executes the first hardware instruction, which computes an intermediate stage of an AI model. The data buffer stores data generated from executing the first hardware instruction. The command circuitry determines that the second hardware instruction is a hardware-initiated store instruction for transferring the data from the data buffer. Based on the hardware-initiated store instruction, the communication circuitry transfers the data from the data buffer to a memory device of a computing system, which includes the mission-critical processor, via a communication interface.

3. 发明申请

WO2019173075A4 MISSION-CRITICAL AI PROCESSOR WITH MULTI-LAYER FAULT TOLERANCE SUPPORT 审中-公开
公开(公告)号：WO2019173075A4
公开(公告)日：2019-09-12
申请号：PCT/US2019/019451
申请日：2019-02-25
申请人： DINOPLUSAI HOLDINGS LIMITED , CHIN, Chung Kuang , HU, Yujie , WU, Tong , GOLD, Clifford , WONG, Yick Kei , WANG, Xiaosong , SERTILLANGE, Steven , ZHU, Zongwei
发明人： CHIN, Chung Kuang , HU, Yujie , WU, Tong , GOLD, Clifford , WONG, Yick Kei , WANG, Xiaosong , SERTILLANGE, Steven , ZHU, Zongwei
IPC分类号： G06N3/08
摘要： Embodiments described herein provide a mission-critical artificial intelligence (AI) processor (MAIP), which includes multiple types of HEs (hardware elements) comprising one or more HEs configured to perform operations associated with multi-layer NN (neural network) processing, at least one spare HE, a data buffer to store correctly computed data in a previous layer of multi-layer NN processing computed, and fault tolerance (FT) control logic. The FT control logic is configured to: determine a fault in a current layer NN processing associated with the HE; cause the correctly computed data in the previous layer of multi-layer NN processing to be copied or moved to said at least one spare HE; and cause said at least one spare HE to perform the current layer NN processing using said at least one spare HE and the correctly computed data in the previous layer of multi-layer NN processing.

4. 发明申请

WO2019173076A1 COMPUTING DEVICE FOR FAST WEIGHTED SUM CALCULATION IN NEURAL NETWORKS 审中-公开
公开(公告)号：WO2019173076A1
公开(公告)日：2019-09-12
申请号：PCT/US2019/019469
申请日：2019-02-25
申请人： DINOPLUSAI HOLDINGS LIMITED , GOLD, Clifford , WU, Tong , HU, Yujie , CHIN, Chung Kuang , WANG, Xiaosong , WONG, Yick Kei
发明人： GOLD, Clifford , WU, Tong , HU, Yujie , CHIN, Chung Kuang , WANG, Xiaosong , WONG, Yick Kei
IPC分类号： G06N3/063 , G06F7/00 , G06F7/499 , G06F7/552 , G06F17/16
摘要： A computing device for fast weighted sum calculation in neural networks is disclosed. The computing device comprises an array of processing elements configured to accept an input array. Each processing element comprises a plurality of multipliers and a multiple levels of accumulators. A set of weights associated with the inputs and a target output are provided to a target processing element to compute the weighted sum for the target output. The device according to the present invention reduces the computation time from M clock cycles to log2M, where M is the size of the input array.

5. 发明申请

WO2019173104A1 AI ACCELERATOR VIRTUALIZATION 审中-公开
公开(公告)号：WO2019173104A1
公开(公告)日：2019-09-12
申请号：PCT/US2019/020074
申请日：2019-02-28
申请人： DINOPLUSAI HOLDINGS LIMITED , HU, Yujie , WANG, Xiaosong , WU, Tong , SERTILLANGE, Steven
发明人： HU, Yujie , WANG, Xiaosong , WU, Tong , SERTILLANGE, Steven
IPC分类号： G06N3/067
摘要： An AI (Artificial Intelligence) processor for Neural Network (NN) Processing shared by multiple users is disclosed. The AI processor comprises a Multiplier Unit (MXU), a Scalar Computing Unit (SCU), a unified buffer coupled to the MXU and SCU to store data and a control circuitry coupled to the CCU and the unified buffer. The MXU comprises a plurality of Processing Elements (PEs) responsible for computing matrix multiplications. The SCU coupled to output of the MXU is responsible for computing the activation function. The control circuitry is configured to perform the space division and time division NN processing for a plurality of users. At one time instance, at least one of the MXU and SCU is shared by two or more users; and at least one user is using a part of the MXU while the other user is using a part of the SCU.

6. 发明申请

WO2023069074A1 METHOD AND APPARATUS OF PROGRAMMABLE QUANTIZATION/DE-QUANTIZATION ENGINE 审中-公开
公开(公告)号：WO2023069074A1
公开(公告)日：2023-04-27
申请号：PCT/US2021/055527
申请日：2021-10-19
申请人： PICOAI LIMITED , SERTILLANGE, Steven , WU, Tong , SABER, Ahmed , HU, Yujie
发明人： SERTILLANGE, Steven , WU, Tong , SABER, Ahmed , HU, Yujie
IPC分类号： H03K19/17724 , H03K19/17728 , H03K19/20 , H03K19/21 , H03K19/23
摘要： A configurable apparatus for quantization and/or de-quantization in artificial intelligence processing is disclosed. The configurable apparatus comprises an input interface configured to receive input data, an output interface configured to provide output data, a multiplier, a shifter, and an adder. The multiplier is configured to multiply data of the first multiplier input and the second multiplier input. The shifter is configured to shift data of the first shifter input according to a shift value. The first shifter input is coupled to the multiplier output. The second shifter input receives the shift value. The first multiplier input is selectively coupled to the input interface or to the adder output, and the second multiplier input is configured to receive a multiplication value. The first adder input is selectively coupled to the input interface or to the shifter output. The output interface is coupled to the shifter output or the adder output selectively.

7. 发明申请

WO2023069063A1 METHOD AND APPARATUS OF EFFICIENT IMAGE DATA ACCESS FOR ARTIFICIAL INTELLIGENCE PROCESSING 审中-公开
公开(公告)号：WO2023069063A1
公开(公告)日：2023-04-27
申请号：PCT/US2021/055378
申请日：2021-10-18
申请人： PICOAI LIMITED , FENG, Yutian , SERTILLANGE, Steven , WU, Tong
发明人： FENG, Yutian , SERTILLANGE, Steven , WU, Tong
IPC分类号： G06T15/20 , G06F12/08
摘要： A sliding window cache apparatus for processing image data and a method for using the apparatus are disclosed. The sliding window cache apparatus comprises a vertical data cache including vertical cache memory to store one or more lines or partial lines of the image data; a horizontal data cache including horizontal cache memory to store one or more horizontal data segments; a cache tag control logic configured to determine whether the vertical data cache is hit or miss and whether the horizontal data cache is hit or miss for one or more AI process instructions involved with memory read from the image data memory. The cache tag control logic provides the cache operation commands to access the cached data in the vertical data cache and the horizontal data cache for a current window prior to the cached data required by a target AI processor to process the current window.

8. 发明申请

WO2012119364A1 Method and apparatus for determining an optical signal-to-noise ratio (OSNR) penalty 审中-公开
标题翻译：用于确定光信噪比（OSNR）罚分的方法和装置
公开(公告)号：WO2012119364A1
公开(公告)日：2012-09-13
申请号：PCT/CN2011/077672
申请日：2011-07-27
申请人： HUAWEI TECHNOLOGIES CO., LTD. , YE, Yabin , WU, Tong , ZHANG, Sen
发明人： YE, Yabin , WU, Tong , ZHANG, Sen
IPC分类号： H04B10/18
CPC分类号： H04B10/07953
摘要： Method and apparatus for determining an optical signal-to-noise ratio (OSNR) penalty. A method for determining an optical signal-to-noise ratio penalty as a measure for a quality of an optical signal transmitted via an optical link between a source optical node and a destination optical node in an optical network, the method comprises collecting (101) information of the optical link; determining (103) a configuration parameter Pconf of the optical link based on the information of the optical link; adjusting (105) the configuration parameter Pconf to an adjusted configuration parameter Pconf according to linear impairments in the optical link; and determining (107) the optical signal-to-noise ratio penalty based on a non-linear function of the adjusted configuration parameter Pconf, the non-linear function accounting for non-linear impairments in the optical link.
摘要翻译：用于确定光信噪比（OSNR）罚分的方法和装置。一种用于确定光信噪比损失的方法，用于通过光网络中的源光节点和目的地光节点之间的光链路传输的光信号的质量的量度，所述方法包括：收集（101）光链路信息; 基于所述光链路的信息确定（103）所述光链路的配置参数Pconf; 根据光链路中的线性损伤将配置参数Pconf调整（105）到调整配置参数Pconf; 以及基于经调整的配置参数Pconf的非线性函数来确定（107）光信噪比损失，所述非线性函数考虑光链路中的非线性损伤。

9. 发明申请

WO2022000470A1 METHOD AND APPARATUS FOR MANAGING INTER-RAT CELL HANDOVER 审中-公开
公开(公告)号：WO2022000470A1
公开(公告)日：2022-01-06
申请号：PCT/CN2020/100139
申请日：2020-07-03
申请人： QUALCOMM INCORPORATED , HUI, ChaoFeng , LENG, Bing , XUE, Liang , WU, Tong , CHEN, Huawen
发明人： HUI, ChaoFeng , LENG, Bing , XUE, Liang , WU, Tong , CHEN, Huawen
IPC分类号： H04W36/00
摘要： The apparatus of wireless communication includes UE and LTE cells that may block a handover from the LTE cell to a 5G NR Cell while an IMS call service is provided via the LTE cell. The UE may block the handover process to the 5G NR cell by blocking a transmission of a measurement report of neighboring cells to the LTE cell or filtering out the measurements of neighboring 5G NR cells in the measurement report of the neighboring cells. The LTE cell may block the initiation of the handover process to the 5G NR cell by excluding a measurement control to trigger a handover to the 5G NR cell in an instruction to modify the RRC connection, or ignoring the handover to the 5G NR cell indicated in the measurement report received from the UE.

10. 发明申请

WO2020131141A1 INTERACTIVE OBJECT HAVING LIGHT-TRANSMISSIVE PATTERN WITH CONTROLLED HOLE-SHAPE 审中-公开
公开(公告)号：WO2020131141A1
公开(公告)日：2020-06-25
申请号：PCT/US2019/014709
申请日：2019-01-23
申请人： GOOGLE LLC , ZARICH, Mark , WU, Tong , ALLEN, Bryan , BEARDSLEY, Brendon , BERNSTEIN, Adam
发明人： ZARICH, Mark , WU, Tong , ALLEN, Bryan , BEARDSLEY, Brendon , BERNSTEIN, Adam
IPC分类号： G06F3/044 , A41D1/00 , F21V33/00
摘要： Systems and methods for interactive objects including illuminated surfaces are provided. An interactive object includes a flexible substrate comprising a soft material that is non-transmissive to light. The interactive object includes a plurality of holes extending from an inside surface of the flexible substrate to an outside surface of the flexible substrate to form a light-transmissive pattern at one or more locations of the flexible substrate. An illuminated surface is positioned adjacent to the inside surface of the flexible substrate with its front surface extending along at least a portion of the inside surface of the flexible substrate at the one or more locations including the plurality of holes. The interactive object includes one or more electronic components electrically coupled to the illuminated surface to control the illuminated surface to provide a visual output.

你已经成功收藏专利！

检索式保存成功!

IPRDB

热门服务

关于我们

友情链接

联系方式