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 2405.01524cs.LGcs.AIcs.CV
收藏

一种基于可分性的泛化量化方法:哪一层表现最佳?

A separability-based approach to quantifying generalization: which layer is best?

Yale University(耶鲁大学)

Luciano Dyballa, Evan Gerritz, Steven W. Zucker

AI总结:

针对深度学习模型泛化能力难以评估、最优泛化层不明确的问题,提出基于中间层潜在嵌入的泛化量化方法,发现高准确率不代表高泛化性、深层并非总泛化最优,揭示了不同层的内在泛化能力。

摘要:

对于深度学习分类模型和基础模型而言,对未见过数据的泛化能力仍缺乏充分理解,在开集场景下尤其如此。我们该如何评估网络依照小样本学习、分布外泛化、域适应和类别发现的思路,适配其输入空间的新变体或扩展版本的能力?网络的哪些层可能泛化效果最好?我们提出了一种新方法,用于评估网络表征采样域的能力,无论网络是否在该域的所有类别上完成训练。我们的方法如下:在特定域上对用于视觉分类的前沿预训练模型进行微调后,我们评估其在该域内相关但有差异的变体数据上的性能。在无监督和有监督设置下,我们将泛化能力量化为中间层生成的未见数据潜在嵌入的函数。通过研究网络的所有阶段,我们发现:(1)高分类准确率并不意味着高泛化性;(2)模型中更深的层并非总能实现最佳泛化,这对模型剪枝具有参考意义。由于在不同数据集上观察到的趋势基本一致,我们得出结论:我们的方法揭示了模型不同层泛化能力的内在属性(的一种函数形式)。我们的代码可在https://github.com/dyballa/generalization获取。

英文摘要:

Generalization to unseen data remains poorly understood for deep learning classification and foundation models, especially in the open set scenario. How can one assess the ability of networks to adapt to new or extended versions of their input space in the spirit of few-shot learning, out-of-distribution generalization, domain adaptation, and category discovery? Which layers of a network are likely to generalize best? We provide a new method for evaluating the capacity of networks to represent a sampled domain, regardless of whether the network has been trained on all classes in that domain. Our approach is the following: after fine-tuning state-of-the-art pre-trained models for visual classification on a particular domain, we assess their performance on data from related but distinct variations in that domain. Generalization power is quantified as a function of the latent embeddings of unseen data from intermediate layers for both unsupervised and supervised settings. Working throughout all stages of the network, we find that (i) high classification accuracy does not imply high generalizability; and (ii) deeper layers in a model do not always generalize the best, which has implications for pruning. Since the trends observed across datasets are largely consistent, we conclude that our approach reveals (a function of) the intrinsic capacity of the different layers of a model to generalize. Our code is available at https://github.com/dyballa/generalization

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。

↑