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 2103.10427cs.LGcs.CV
收藏

深度网络中的低秩简单性偏差

The Low-Rank Simplicity Bias in Deep Networks

MIT CSAIL(麻省理工学院计算机科学与人工智能实验室);Google Research(谷歌研究院);Adobe Research(奥多比研究院)

Minyoung Huh, Hossein Mobahi, Richard Zhang, Brian Cheung, Pulkit Agrawal, Phillip Isola

摘要:

现代深度神经网络与它们所训练的数据相比是高度过度参数化的,但它们通常具有非常好的泛化能力。最近的一系列研究提出了这样一个问题:为什么深度网络不会过度拟合其训练数据?在这项工作中,我们做了一系列的实证观察,调查和扩展的假设,更深的网络是归纳偏向于找到解决方案与较低的有效秩嵌入。我们推测,这种偏见的存在,因为映射到低有效秩嵌入的函数的量随着深度的增加而增加。我们的经验表明,我们的主张适用于有限宽度的线性和非线性模型的实际学习范式,并表明,在自然数据,这些往往是解决方案,推广良好。然后,我们证明了简单性偏差在初始化和训练后都存在,并且对超参数和学习方法具有弹性。我们进一步展示了如何使用深度非线性模型的线性过参数化来诱导低秩偏差,从而在不改变建模能力的情况下提高CIFAR和ImageNet的泛化性能。

英文摘要:

Modern deep neural networks are highly over-parameterized compared to the data on which they are trained, yet they often generalize remarkably well. A flurry of recent work has asked: why do deep networks not overfit to their training data? In this work, we make a series of empirical observations that investigate and extend the hypothesis that deeper networks are inductively biased to find solutions with lower effective rank embeddings. We conjecture that this bias exists because the volume of functions that maps to low effective rank embedding increases with depth. We show empirically that our claim holds true on finite width linear and non-linear models on practical learning paradigms and show that on natural data, these are often the solutions that generalize well. We then show that the simplicity bias exists at both initialization and after training and is resilient to hyper-parameters and learning methods. We further demonstrate how linear over-parameterization of deep non-linear models can be used to induce low-rank bias, improving generalization performance on CIFAR and ImageNet without changing the modeling capacity.

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。

↑