-
公开(公告)号：WO2022216878A1
公开(公告)日：2022-10-13
申请号：PCT/US2022/023739
申请日：2022-04-06
申请人： GOOGLE LLC
摘要： Methods, systems, and apparatus, including computer programs encoded on computer storage media, for determining a hardware datapath for a hardware accelerator computer chip.
-
公开(公告)号：WO2022134001A1
公开(公告)日：2022-06-30
申请号：PCT/CN2020/139384
申请日：2020-12-25
申请人： 深圳晶泰科技有限公司
摘要： 本发明提供一种基于容器化技术的机器学习模型框架的开发方法与系统，包括如下步骤：机器学习模型开发：建立对应的机器学习模型，通过编程的方式构建模型并将对应的代码提交到代码仓库中；模型交互式测试：挑选合适的测试数据对已开发的机器学习模型进行验证；模型参数优化与监控：已经调试验证成功的机器学习模型及参数，可以被其他使用者直接调用，通过代码层面的参数调整来实现最优化的机器学习模型训练与结果监控，最终得到一个可以供部署使用的机器学习模型及参数。本发明通过容器化技术实现跨平台的可扩展的部署实施能力。
-
公开(公告)号：WO2022090809A1
公开(公告)日：2022-05-05
申请号：PCT/IB2021/052653
申请日：2021-03-30
IPC分类号： G06N3/10
摘要： Methods and system of machine learning life cycle management for networks are described. A first machine learning (ML) model is generated in a development platform based on a first set of training data. The first ML model is deployed for use in a first network. The network is operated separately from the development platform. Data resulting from the use of the first ML model in the network is received through an open communication interface from the network. The first ML model is updated based on the received data to obtain a second ML model for deployment in the network.
-
公开(公告)号：WO2022082836A1
公开(公告)日：2022-04-28
申请号：PCT/CN2020/124656
申请日：2020-10-29
申请人： 南京博芯电子技术有限公司
发明人： 王镇
摘要： 一种面向稀疏递归神经网络的均衡运算加速方法与系统，根据权重矩阵的稀疏性的仲裁结果确定调度信息，选择工作电压和运行频率与调度信息相匹配的运算子模块或工作电压和运行频率经过调整后与调度信息相匹配的运算子模块，采用选定的运算子模块依次进行跳零运算和乘加运算，实现均衡运算加速。均衡运算加速系统包括数据传输模块、内置多个相互独立的运算子模块的均衡运算调度模块、电压可调均衡运算模块。配置错误监视器实现电压动态调整；数据传输模块快速交换读/写存储器的配置，减少额外数据传输；在提升运算速度的情况下，通过均衡调度的方式来减少功耗，并减少计算时电压的波动。
-
公开(公告)号：WO2021259482A1
公开(公告)日：2021-12-30
申请号：PCT/EP2020/067800
申请日：2020-06-25
IPC分类号： G06N3/063 , G06N3/04 , G06N3/10 , G06F30/30 , G06N3/0445 , G06N3/0454 , G06N3/0481 , G06N3/0635 , G06N3/084
摘要： Systems and methods are provided for analog hardware realization of neural networks. The method incudes obtaining a neural network topology and weights of a trained neural network. The method also includes transforming the neural network topology to an equivalent analog network of analog components. The method also includes computing a weight matrix for the equivalent analog network based on the weights of the trained neural network. Each element of the weight matrix represents a respective connection between analog components of the equivalent analog network. The method also includes generating a schematic model for implementing the equivalent analog network based on the weight matrix, including selecting component values for the analog components.
-
公开(公告)号：WO2021235312A1
公开(公告)日：2021-11-25
申请号：PCT/JP2021/018193
申请日：2021-05-13
申请人： ソニーグループ株式会社
发明人： 佐宗 馨
IPC分类号： G06N3/10
摘要： 情報処理装置（１００）は、ニューラルネットワークを用いた推論器による推論結果を利用する情報処理システムに適用される情報処理装置であって、取得部（１１３）と、算出部（１１４）と、決定部（１１５）とを備える。取得部（１１３）は、情報処理システムの全体のリソース使用量を取得する。算出部（１１４）は、リソース使用量に基づいて、推論器による推論処理の少なくとも一部の計算に割り当てる目標リソース使用量を算出する。決定部（１１５）は、目標リソース使用量に対応する計算方法を決定する。
-
公开(公告)号：WO2021125794A1
公开(公告)日：2021-06-24
申请号：PCT/KR2020/018460
申请日：2020-12-16
发明人： KIM, Dongwan
摘要： Provided are a computing device and a method of operating the same. The computing device may include a memory storing one or more instructions, and a processor configured to execute the one or more instructions stored in the memory to recognize a trigger word for another voice assistant, and analyze a response of the other voice assistant to a speech given to the other voice assistant. The computing device simulates an operation performed to correspond to the speech, and performs a subsequent operation, based on a result of the simulating and a result of the analyzing of the response of the other voice assistant.
-
公开(公告)号：WO2021123837A2
公开(公告)日：2021-06-24
申请号：PCT/GB2020/053331
申请日：2020-12-21
申请人： ARM LIMITED
发明人： HOPKINS, Andrew Brian Thomas , INGRAM, Graeme Leslie , ROSEMARINE, Elliot Maurice Simon , PRIORE, Antonio
IPC分类号： G06N3/063 , G06N3/04 , G06F11/14 , G06F11/16 , G06F11/1476 , G06F11/1494 , G06F11/1497 , G06F11/1629 , G06N3/0454 , G06N3/10
摘要： A method of performing fault detection during computations relating to a neural network comprising a first neural network layer and a second neural network layer in a data processing system, the method comprising: scheduling computations onto data processing resources for the execution of the first neural network layer and the second neural network layer, wherein the scheduling includes: for a given one of the first neural network layer and the second neural network layer, scheduling a respective given one of a first computation and a second computation as a non-duplicated computation, in which the given computation is at least initially scheduled to be performed only once during the execution of the given neural network layer; and for the other of the first and second neural network layers, scheduling the respective other of the first and second computations as a duplicated computation, in which the other computation is at least initially scheduled to be performed at least twice during the execution of the other neural network layer to provide a plurality of outputs; performing computations in the data processing resources in accordance with the scheduling; and comparing the outputs from the duplicated computation to selectively provide a fault detection operation during processing of the other neural network layer.
-
公开(公告)号：WO2021079763A1
公开(公告)日：2021-04-29
申请号：PCT/JP2020/038428
申请日：2020-10-12
申请人： ソニー株式会社
发明人： 大渕 愉希夫
IPC分类号： G06F9/50 , G06N3/10 , G06F3/0481
摘要： 本技術の一形態に係る情報処理方法は、コンピュータシステムに以下のステップを実行させる。ニューラルネットワークを構成する複数のレイヤを分割して複数のハードウェアに実行させる分割処理に必要な処理リソースを算出するステップ。前記分割処理における前記複数のレイヤと前記複数のハードウェアとの対応関係とともに、前記処理リソースを示すＧＵＩ画面を生成するステップ。
-
公开(公告)号：WO2021051006A1
公开(公告)日：2021-03-18
申请号：PCT/US2020/050569
申请日：2020-09-11
申请人： VIANAI SYSTEMS, INC.
摘要： One embodiment of the present invention sets forth a technique for creating a machine learning model. The technique includes generating a user interface comprising one or more components for visually generating the machine learning model. The technique also includes modifying source code specifying a plurality of mathematical expressions that define the machine learning model based on user input received through the user interface. The technique further includes compiling the source code into compiled code that, when executed, causes one or more parameters of the machine learning model to be learned during training of the machine learning model.
-
-
-
-
-
-
-
-
-
搜索结果
209 条记录 (耗时 0.025 秒)
