arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2103.13558cs.LGcs.AIcs.CV
收藏

用于判别式和生成式持续学习的有效特征变换

Efficient Feature Transformations for Discriminative and Generative Continual Learning

Duke University(杜克大学);Facebook AI(脸书人工智能研究院);IIT Kanpur(坎普尔印度理工学院)

Vinay Kumar Verma, Kevin J Liang, Nikhil Mehta, Piyush Rai, Lawrence Carin

摘要:

随着神经网络越来越多地应用于现实世界的应用,解决分布转移和顺序任务学习而不遗忘的机制至关重要。结合网络扩展的方法通过自然地增加用于学习新任务的模型容量,同时避免灾难性遗忘而显示出希望。然而,这些类型的方法中的许多方法的附加参数的数量的增长在较大规模上可能在计算上是昂贵的,有时是令人望而却步的。相反,我们提出了一个简单的特定于任务的持续学习,我们称之为有效的特征转换(EFT)的特征映射转换策略。这些EFT为学习新任务提供了强大的灵活性,只需向基础架构添加最少的参数即可实现。我们进一步提出了一个特征距离最大化策略,它显着提高了类增量设置中的任务预测,而不需要昂贵的生成模型。我们通过一组广泛的实验来证明我们的方法在区分性(CIFAR-100和ImageNet-1 K)和生成性(LSUN,CUB-200,Cats)任务序列中的有效性和效率。即使是低个位数的参数增长率,EFT也可以在广泛的设置中胜过许多其他持续学习方法。

英文摘要:

As neural networks are increasingly being applied to real-world applications, mechanisms to address distributional shift and sequential task learning without forgetting are critical. Methods incorporating network expansion have shown promise by naturally adding model capacity for learning new tasks while simultaneously avoiding catastrophic forgetting. However, the growth in the number of additional parameters of many of these types of methods can be computationally expensive at larger scales, at times prohibitively so. Instead, we propose a simple task-specific feature map transformation strategy for continual learning, which we call Efficient Feature Transformations (EFTs). These EFTs provide powerful flexibility for learning new tasks, achieved with minimal parameters added to the base architecture. We further propose a feature distance maximization strategy, which significantly improves task prediction in class incremental settings, without needing expensive generative models. We demonstrate the efficacy and efficiency of our method with an extensive set of experiments in discriminative (CIFAR-100 and ImageNet-1K) and generative (LSUN, CUB-200, Cats) sequences of tasks. Even with low single-digit parameter growth rates, EFTs can outperform many other continual learning methods in a wide range of settings.

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。

↑