Continual learning: a feature extraction formalization, an efﬁcient algorithm, and barriers-猫眼课题宝

DOI：

发表时间：

期刊：

影响因子：

通讯作者：

中科院分区：

其他

文献类型：

作者：

Binghui Peng;Andrej Risteski

持续学习是机器学习中的一种新兴范式，其中模型以在线方式暴露于来自多个不同分布（即环境）的数据，并且预计会适应分布变化。准确地说，目标是在新环境中表现良好，同时保持以前环境的性能（即避免“灾难性遗忘”）。虽然这种设置在应用社区中受到了很多关注，但还没有理论上的工作，甚至还没有正式确定所需的保证。在本文中，我们提出了一个通过特征提取框架进行持续学习的框架，即在每个环境中训练特征和分类器的框架。当特征是线性的时，我们设计了一个有效的基于梯度的算法DPGrad，保证在当前环境下表现良好，并避免灾难性遗忘。在一般情况下，当特征是非线性的，我们证明这样的算法不存在，无论是否有效。

Continual learning is an emerging paradigm in machine learning, wherein a model is exposed in an online fashion to data from multiple different distributions (i.e. environments), and is expected to adapt to the distribution change. Precisely, the goal is to perform well in the new environment, while simultaneously retaining the performance on the previous environments (i.e. avoid “catastrophic forgetting”). While this setup has enjoyed a lot of attention in the applied community, there hasn’t be theoretical work that even formalizes the desired guarantees. In this paper, we propose a framework for continual learning through the framework of feature extraction—namely, one in which features, as well as a classiﬁer, are being trained with each environment. When the features are linear, we design an efﬁcient gradient-based algorithm DPGrad , that is guaranteed to perform well on the current environment, as well as avoid catastrophic forgetting. In the general case, when the features are non-linear, we show such an algorithm cannot exist, whether efﬁcient or not.

Continual learning: a feature extraction formalization, an efﬁcient algorithm, and barriers

文献摘要

相似文献