arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2205.05967cs.CV
收藏

面向高效知识迁移的目标感知网络架构搜索与压缩

Target Aware Network Architecture Search and Compression for Efficient Knowledge Transfer

RV University（RV大学）；Google Research（谷歌研究院）；New Jersey Institute of Technology（新泽西理工学院）；Indian Institute of Information Technology Allahabad（印度信息技术学院阿拉哈巴德分校）

S. H. Shabbeer Basha, Debapriya Tula, Sravan Kumar Vinakota, Shiv Ram Dubey

备注：
This paper is accepted for publication in Multimedia Systems Journal
阅读 PDF
arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出两阶段框架 TASCNet，通过自动搜索目标任务适配的深层结构并剪除冗余滤波器，在保持性能的同时减少参数与 FLOPs，实现高效知识迁移。

摘要：

迁移学习使卷积神经网络（CNN）能够从源域获取知识，并将其迁移到目标域；在目标域中，收集大规模带标注样本既耗时又昂贵。按照常规做法，在将从一个任务学到的知识迁移到另一个任务时，会在目标数据集上对预训练 CNN 的较深层进行微调。然而，这些层最初是为源任务设计的，对于目标任务而言可能存在过度参数化。因此，由于网络复杂度较高，在目标数据集上微调这些层可能会影响 CNN 的泛化能力。为解决这一问题，我们提出了一个名为 TASCNet 的两阶段框架，以实现高效的知识迁移。在第一阶段，自动学习较深层的配置，并在目标数据集上进行微调。随后，在第二阶段，从经过微调的 CNN 中剪除冗余滤波器，以降低网络针对目标任务的复杂度，同时保持性能。这种两阶段机制从假设空间中找到预训练 CNN 的紧凑版本，使其具有最优结构（卷积层中的滤波器数量、密集层中的神经元数量等）。所提方法的有效性通过在 CalTech-101、CalTech-256 和 Stanford Dogs 数据集上使用 VGG-16、ResNet-50 和 DenseNet-121 进行评估。与计算机视觉任务类似，我们还在 Movie Review 情感分析任务上进行了实验。所提出的 TASCNet 通过同时减少可训练参数和 FLOPs，降低了预训练 CNN 在目标任务上的计算复杂度，从而实现资源高效的知识迁移。源代码可在 https://github.com/Debapriya-Tula/TASCNet 获取。

英文摘要：

Transfer Learning enables Convolutional Neural Networks (CNN) to acquire knowledge from a source domain and transfer it to a target domain, where collecting large-scale annotated examples is time-consuming and expensive. Conventionally, while transferring the knowledge learned from one task to another task, the deeper layers of a pre-trained CNN are finetuned over the target dataset. However, these layers are originally designed for the source task which may be over-parameterized for the target task. Thus, finetuning these layers over the target dataset may affect the generalization ability of the CNN due to high network complexity. To tackle this problem, we propose a two-stage framework called TASCNet which enables efficient knowledge transfer. In the first stage, the configuration of the deeper layers is learned automatically and finetuned over the target dataset. Later, in the second stage, the redundant filters are pruned from the fine-tuned CNN to decrease the network's complexity for the target task while preserving the performance. This two-stage mechanism finds a compact version of the pre-trained CNN with optimal structure (number of filters in a convolutional layer, number of neurons in a dense layer, and so on) from the hypothesis space. The efficacy of the proposed method is evaluated using VGG-16, ResNet-50, and DenseNet-121 on CalTech-101, CalTech-256, and Stanford Dogs datasets. Similar to computer vision tasks, we have also conducted experiments on Movie Review Sentiment Analysis task. The proposed TASCNet reduces the computational complexity of pre-trained CNNs over the target task by reducing both trainable parameters and FLOPs which enables resource-efficient knowledge transfer. The source code is available at: https://github.com/Debapriya-Tula/TASCNet.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。

↑