使用用于深度神经网络的改进的训练和学习的方法和系统
By introducing communication coupling between the GPU and the host processor core and the SIMT architecture in the graphics processing unit, the graphics and machine learning operations are optimized, solving the problem of computational intensity in deep convolutional neural networks, improving training and learning efficiency, and reducing resource requirements.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- INTEL CORP
- Filing Date
- 2017-04-07
- Publication Date
- 2026-07-17
AI Technical Summary
Existing graphics processors are computationally intensive when processing deep convolutional neural networks. The training and learning process requires a large amount of data and computing resources, leading to limitations in the computing power and memory of computing devices.
By communicating and coupling the graphics processing unit (GPU) with the host processor core, commands are processed efficiently using dedicated circuit logic, optimizing graphics and machine learning operations, and combining a single instruction multithreaded architecture (SIMT) to process graphics data and machine learning tasks in parallel.
It improves the training and learning efficiency of deep convolutional neural networks, reduces the demand for computing and storage resources, and supports the processing of larger datasets.
Smart Images
Figure CN110352432B_ABST