.png){kind=icon}
返回
Memory-Replay Knowledge Distillation
.png){kind=icon}
OA
AI
DOI：10.3390/s21082792.png)
.png){kind=icon}
.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Knowledge Distillation (KD), which transfers the knowledge from a teacher to a student network by penalizing their Kullback-Leibler (KL) divergence, is a widely used tool for Deep Neural Network (DNN) compression in intelligent sensor systems. Traditional KD uses pre-trained teacher, while self-KD distills its own knowledge to achieve better performance. The role of the teacher in self-KD is usually played by multi-branch peers or the identical sample with different augmentation. However, the mentioned self-KD methods above have their limitation for widespread use. The former needs to redesign the DNN for different tasks, and the latter relies on the effectiveness of the augmentation method. To avoid the limitation above, we propose a new self-KD method, Memory-replay Knowledge Distillation (MrKD), that uses the historical models as teachers. Firstly, we propose a novel self-KD training method that penalizes the KD loss between the current model's output distributions and its backup outputs on the training trajectory. This strategy can regularize the model with its historical output distribution space to stabilize the learning. Secondly, a simple Fully Connected Network (FCN) is applied to ensemble the historical teacher's output for a better guidance. Finally, to ensure the teacher outputs offer the right class as ground truth, we correct the teacher logit output by the Knowledge Adjustment (KA) method. Experiments on the image (dataset CIFAR-100, CIFAR-10, and CINIC-10) and audio (dataset DCASE) classification tasks show that MrKD improves single model training and working efficiently across different datasets. In contrast to the existing fancy self-KD methods with various external knowledge, the effectiveness of MrKD sheds light on the usually abandoned historical models during the training trajectory.
Keyword:
Deep Neural Network
self-knowledge distillation
training trajectory
Knowledge Adjustment
Fully Connected Network
image classification
audio classification
AI总结
对已上传原文的论文进行重点信息的提取，主要内容包括：简要概述、研究摘要、背景介绍、关键亮点、图文解析、展望与总结。
期刊
IF:
3.5
论文数:
7.2W
被引数:
20.9W
机构
引用论文
.png){kind=icon}
Study of bi-directional buck-boost converter topologies for application in electrical vehicle motor drives应用于电动汽车电机驱动的双向buck-boost变换器拓扑研究
PREAI
分享收藏Synthesis of n-type semiconducting diamond film using diphosphorus pentaoxide as the doping source以五氧化二磷为掺杂源合成n型半导体金刚石膜
PREAI
分享收藏The power of obfuscation techniques in malicious JavaScript code: A measurement study恶意JavaScript代码中混淆技术的功能: 一项测量研究OAAI
OAAI分享收藏
查看更多内容