-
公开(公告)号：WO2023090502A1
公开(公告)日：2023-05-25
申请号：PCT/KR2021/017368
申请日：2021-11-24
Applicant: 서울대학교산학협력단
Abstract: 본 발명은 프레임 양자화에 기반한 분산 행렬 곱 연산 방법 및 장치에 관한 것으로, 프레임 양자화에 기반한 부호화 컴퓨팅을 이용하여 복수의 연산 노드에서 분산 행렬 곱 연산을 수행하는 방법 및 장치를 제공한다. 이로써 고차원 행렬 곱 처리 성능이 개선된다.
-
2.发明申请
公开(公告)号：WO2023089610A1
公开(公告)日：2023-05-25
申请号：PCT/IL2022/051224
申请日：2022-11-17
Applicant: DECI.AI LTD.
Inventor： TROPP, Oren
Abstract: Embodiments of the invention may include a system and method of automatically optimizing calculation of a butterfly transform by a processing unit. The processing unit may be adapted to perform atomic [NxN] matrix-matrix multiplication operations. Embodiments of the invention may include: receiving an input data matrix of dimensions [MxB], representing a batch of B input data vectors, each of length M; arranging the input data matrix into S section matrices of dimensions [N rows x K columns], wherein K>= N and K
-
公开(公告)号：WO2023080320A1
公开(公告)日：2023-05-11
申请号：PCT/KR2021/017878
申请日：2021-11-30
Applicant: 주식회사 토이코스
Inventor： 엄준석
IPC: G06Q50/06 , G06Q50/10 , G01D4/00 , G01F15/06 , G01M3/26 , G06F17/16 , G06N20/00 , G06F30/20 , G06F113/08
Abstract: 본 발명은 다차원 데이터를 이용한 옥내 누수 탐지 및 유형 구분 장치, 방법 및 프로그램을 공개한다. 이 방법은 (a) 제어부가 원격검침으로 측정된 물 사용량 데이터를 인가받아, 수용가별로 최소 물 사용량의 제1 및 제2 측정된 시간 구간을 계산하고, 정상 사용 기계학습부가 상기 물 사용량 데이터에 대하여 제1 기계학습하는 단계; (b) 상기 제어부가 상기 계산된 시간 구간의 크기에 따라 누수 발생의 경우 '옥내 누수 상태'로 라벨링하고, 앙상블 기계학습부가 상기 누수 발생에 대하여 다차원 데이터를 인가받아 제2 기계학습하는 단계; 및 (c) 상기 제어부가 '정상 상태'와 '옥내 누수 상태'의 조합 데이터를 생성하고, '옥내 누수 상태'로 탐지된 경우 누수 유형의 구분 및 누수량 산출을 수행하는 단계; 를 포함하는 것을 특징으로 한다.
-
公开(公告)号：WO2023077770A1
公开(公告)日：2023-05-11
申请号：PCT/CN2022/092990
申请日：2022-05-16
Applicant: 海光信息技术股份有限公司
IPC: G06F17/16
Abstract: 本公开提供了一种应用于矩阵运算的数据处理方法、装置、设备和存储介质。该数据处理方法包括：确定存储有第一矩阵的至少一个第一向量寄存器和存储有第二矩阵的至少一个第二向量寄存器，第一矩阵包括多个第一操作数据，第二矩阵包括多个第二操作数据；获取第一数据选择信息和第二数据选择信息；基于第一数据选择信息，从多个第一操作数据中选择至少一个第一操作数据，以得到至少一个第一目标操作数据；基于第二数据选择信息，从多个第二操作数据中选择至少一个第二操作数据，以得到至少一个第二目标操作数据。该数据处理方法充分利用了矩阵特性，在线程之间有效地复用数据，大大减少了数据的读取次数，降低了功耗。
-
公开(公告)号：WO2023071658A1
公开(公告)日：2023-05-04
申请号：PCT/CN2022/121335
申请日：2022-09-26
Applicant: 华为技术有限公司
Abstract: 本申请公开了人工智能领域中的一种AI模型的处理方法、运算方法及装置，在该处理方法中，以权重块作为稀疏粒度对初始AI模型进行稀疏处理，得到目标AI模型，权重块的尺寸为加速器的最小计算单元的整数倍。本申请的方案减少了计算量，提高了模型的运行速度，减少了运算开销。
-
Patent Agency Ranking
公开(公告)号：WO2023057836A1
公开(公告)日：2023-04-13
申请号：PCT/IB2022/058399
申请日：2022-09-07
Applicant: INTERNATIONAL BUSINESS MACHINES CORPORATION , IBM ISRAEL SCIENCE AND TECHNOLOGY LTD. , IBM (CHINA) INVESTMENT COMPANY LTD.
Inventor： LAVI, Ofer , RONEN, Inbal , RABINOVICH, Ella , ANABY - TAVOR, Ateret , BOAZ, David , SHLOMOV, Segev , AMID, David
IPC: G06F17/16 , G06F40/35 , G10L15/10 , G10L15/1815
Abstract: Automatic measurement of semantic textual similarity of conversations, by: receiving two conversation texts, each comprising a sequence of utterances; encoding each of the sequences of utterances into a corresponding sequence of semantic representations; computing a minimal edit distance between the sequences of semantic representations; and, based on the computation of the minimal edit distance, performing at least one of: quantifying a semantic similarity between the two conversation texts, and outputting an alignment of the two sequences of utterances with each other.
-
公开(公告)号：WO2023046216A1
公开(公告)日：2023-03-30
申请号：PCT/CN2022/132598
申请日：2022-11-17
Applicant: 山东大学
IPC: G06F17/16
Abstract: 本发明属于色谱分析技术领域，涉及一种近红外光谱与特征图谱的图谱转换方法及其应用，包括以下步骤：首先，将光谱原始矩阵X 1和特征图谱原始矩阵X 2进行异常值剔除及预处理，然后进行奇异值分解，在保留相同的主成分数下，得到X 1的得分矩阵S 1和X 2的S 2；将S 1和S 2两个矩阵进行关联；通过公式X 2trans＝[X 1V 1(P 1 T) +(P 2 T)]V 2 T，将转换后的图谱校正为适合从仪器的特征图谱；其中，X 2trans表示转移后的特征图谱矩阵；V 1的含义是X 1的负载矩阵；V 2的含义是X 2的负载矩阵；P 1 T和P 2 T是Ps的两个子矩阵，Ps为S comb＝[S 1,S 2]组合矩阵的负载矩阵；上标"T"表示转置。该方法能够实现近红外光谱和特征(或指纹)图谱之间的图谱转换，实现不同类型仪器之间的图谱转换。
-
公开(公告)号：WO2023043712A1
公开(公告)日：2023-03-23
申请号：PCT/US2022/043289
申请日：2022-09-13
Inventor： YOUNGBLOOD, Nathan
Abstract: A device for performing vector operations is provided. The device includes a photonic crossbar array. The photonic crossbar array includes a plurality of unit cells. One or more of the plurality of unit cells includes a beam splitter, a first photodetector, and a second photodetector. The one or more unit cells are configured to output, as a unit cell output, a third output of the optical signal and a fourth output of the optical signal. The device includes a controller configured to encode a first vector in time-varying amplitudes or time-varying phases of a first electric field, encode a second vector in time-varying amplitudes or time-varying phases of a second electric field, and determine a result of multiplication of the first vector and the second vector based on the unit cell output from the one or more of the plurality of unit cells.
-
公开(公告)号：WO2023037093A1
公开(公告)日：2023-03-16
申请号：PCT/GB2022/052164
申请日：2022-08-22
Applicant: ARM LIMITED
Inventor： WHATMOUGH, Paul Nicholas , LIU, Zhi-Gang , MATTINA, Matthew
Abstract: A matrix multiplication system and method are provided. The system includes a memory that stores one or more weight tensors, a processor and a matrix multiply accelerator (MMA). The processor converts each weight tensor into an encoded block set that is stored in the memory. Each encoded block set includes a number of encoded blocks, and each encoded block includes a data field and an index field. The MMA converts each encoded block set into a reconstructed weight tensor, and convolves each reconstructed weight tensor and an input data tensor to generate an output data matrix.
-
公开(公告)号：WO2023030163A1
公开(公告)日：2023-03-09
申请号：PCT/CN2022/114853
申请日：2022-08-25
Applicant: 咪咕文化科技有限公司 , 中国移动通信集团有限公司
Abstract: 本发明公开了一种三维模型纹理贴图的转换方法、装置、设备以及介质，所述方法包括：根据三维模型中多个初始关键点坐标确定目标平面，所述三维模型分为多个面片，所述初始关键点为所述面片的边界顶点；获取每个所述初始关键点对应的关键点集合，每个所述初始关键点对应的关键点集合包括所述初始关键点坐标以及所述初始关键点的邻域内的初始关键点；根据每个所述关键点集合确定每个所述初始关键点在所述目标平面上的拉普拉斯坐标；根据预设算法以及每个所述初始关键点的所述拉普拉斯坐标将每个所述面片对应的纹理贴图插入所述目标平面，得到二维的纹理贴图。本发明提高了展开的三维模型的纹理贴图准确度。
-
-
-
-
-
-
-
-
-
Search Results
883
records
(took 0.093 seconds)
