Knowledge Commons of Institute of Automation,CAS
Towards Convolutional Neural Networks Compression via Global&Progressive Product Quantization | |
Weihan Chen1,2; Peisong Wang1; Jian Cheng1 | |
2020-09 | |
会议名称 | British Machine Vision Conference (BMVC) |
会议日期 | 2020-9-7 |
会议地点 | 线上举办 |
摘要 | In recent years, we have witnessed the great success of convolutional neural networks in a wide range of visual applications. However, these networks are typically deficient due to the high cost in storage and computation, which prohibits their further extensions to resource-limited applications. In this paper, we introduce Global&Progressive Product Quantization (G&P PQ), an end-to-end product quantization based network compression method, to merge the separate quantization and finetuning process into a consistent training framework. Compared to existing two-stage methods, we avoid the timeconsuming process of choosing layer-wise finetuning hyperparameters and also make the network capable of learning complex dependencies among layers by quantizing globally and progressively. To validate the effectiveness, we benchmark G&P PQ by applying it to ResNet-like architectures for image classification and demonstrate state-of-the-art tradeoff in terms of model size vs. accuracy under extensive compression configurations compared to previous methods. |
收录类别 | 其他 |
语种 | 英语 |
七大方向——子方向分类 | AI芯片与智能计算 |
国重实验室规划方向分类 | 智能计算与学习 |
是否有论文关联数据集需要存交 | 否 |
文献类型 | 会议论文 |
条目标识符 | http://ir.ia.ac.cn/handle/173211/40618 |
专题 | 复杂系统认知与决策实验室_高效智能计算与学习 |
通讯作者 | Jian Cheng |
作者单位 | 1.NLPR & AIRIA, Institute of Automation, Chinese Academy of Sciences 2.School of Artificial Intelligence, University of Chinese Academy of Sciences |
第一作者单位 | 模式识别国家重点实验室 |
通讯作者单位 | 模式识别国家重点实验室 |
推荐引用方式 GB/T 7714 | Weihan Chen,Peisong Wang,Jian Cheng. Towards Convolutional Neural Networks Compression via Global&Progressive Product Quantization[C],2020. |
条目包含的文件 | 下载所有文件 | |||||
文件名称/大小 | 文献类型 | 版本类型 | 开放类型 | 使用许可 | ||
bmvc_final.pdf(336KB) | 会议论文 | 开放获取 | CC BY-NC-SA | 浏览 下载 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论