深度概率网络的高效量化方法

	深度概率网络的高效量化方法
翻译题名	HIGH-EFFICIENT QUANTIZATION METHOD FOR DEEP PROBABILISTIC NETWORK
申请号	US18387463
	2024-07-04
公开（公告）号	US20240220770A1
公开日期	2024-07-04
摘要	一种深度概率网络的高效量化方法，通过混合量化、结构重构和类型优化，达到了良好的效果。首先，对于有向无环图(DAG)结构，对DAG中的所有节点进行聚类，并根据聚类类别对每个节点进行特定算法类型的量化，得到初步量化的深度概率网络。其次，基于输入权重对初步量化的深度概率网络中的多入节点进行重构，结构重构将多入节点转换为仅包含两个输入节点的二叉树网络，并对重构后的结构进行参数重构。最后，采用基于功耗分析和网络精度分析的算法类型搜索方法对所有节点的算法类型进行优化。该方法能够在保持深度概率网络模型精度的同时，显著降低计算复杂度和计算能耗。
翻译摘要	A high-efficient quantization method for a deep probabilistic network achieves good result through hybrid quantization, structure reformulation, and type optimization. Firstly, for a directed acyclic graph (DAG) structure, all nodes in the DAG are clustered, and each node is quantized by a specific arithmetic type based on the clustering category, to obtain a preliminarily quantized deep probabilistic network. Secondly, the multi-in nodes in a preliminarily quantized deep probabilistic network are reformulated based on the input weights, structural reformulation converts a multi-in node into a binary tree network containing only two-input nodes, and parametrical reformulation is performed on the reformulated structure. Finally, arithmetic types of all nodes are optimized by using an arithmetic type search method based on power consumption analysis and network accuracy analysis. The method can significantly reduce computational complexity and energy consumption for computing while maintaining model accuracy of the deep probabilistic network.
当前权利人	Shanghaitech University
专利申请人	Shanghaitech University
	Shen Zhang; Xinzhe Liu; Yajun Ha
公开国别	美国
公开国别简称	US
IPC 分类号	G06N3//04; G06N5//04; G06N7//01
CPC分类号	G06N3//04; G06N5//04; G06N7//01
专利有效性	审中
专利类型	发明申请
专利类型字典	1
当前法律状态	实质审查
简单同族	WO2024138906A1; CN115860126A; US20240220770A1
扩展同族	WO2024138906A1; CN115860126A; US20240220770A1
INPADOC 同族	US20240220770A1
文献类型	专利
条目标识符	https://kms.shanghaitech.edu.cn/handle/2MSLDSTB/396005
专题	信息科学与技术学院_PI研究组_哈亚军组信息科学与技术学院_博士生
作者单位	Shanghaitech University
推荐引用方式 GB/T 7714	Shen Zhang,Xinzhe Liu,Yajun Ha. 深度概率网络的高效量化方法. US18387463[P]. 2024-07-04.