[论文解读] Interpreting Deep Learning: The Machine Learning Rorschach Test?

Adam S. Charles|arXiv (Cornell University)|Jun 1, 2018

Advanced Statistical Modeling Techniques参考文献 90被引用 8

一句话总结

本文通过将深度神经网络（DNNs）视为一种‘机器学习罗夏测验’，研究其可解释性，其中模型行为受到主观解释的影响。文章指出，尽管DNNs在实践中取得了经验上的成功，但其内部机制仍理解不足，尤其是在泛化能力、对抗鲁棒性以及优化作用方面，凸显了实际性能与理论洞察之间的关键差距。

ABSTRACT

Theoretical understanding of deep learning is one of the most important tasks facing the statistics and machine learning communities. While deep neural networks (DNNs) originated as engineering methods and models of biological networks in neuroscience and psychology, they have quickly become a centerpiece of the machine learning toolbox. Unfortunately, DNN adoption powered by recent successes combined with the open-source nature of the machine learning community, has outpaced our theoretical understanding. We cannot reliably identify when and why DNNs will make mistakes. In some applications like text translation these mistakes may be comical and provide for fun fodder in research talks, a single error can be very costly in tasks like medical imaging. As we utilize DNNs in increasingly sensitive applications, a better understanding of their properties is thus imperative. Recent advances in DNN theory are numerous and include many different sources of intuition, such as learning theory, sparse signal analysis, physics, chemistry, and psychology. An interesting pattern begins to emerge in the breadth of possible interpretations. The seemingly limitless approaches are mostly constrained by the lens with which the mathematical operations are viewed. Ultimately, the interpretation of DNNs appears to mimic a type of Rorschach test --- a psychological test wherein subjects interpret a series of seemingly ambiguous ink-blots. Validation for DNN theory requires a convergence of the literature. We must distinguish between universal results that are invariant to the analysis perspective and those that are specific to a particular network configuration. Simultaneously we must deal with the fact that many standard statistical tools for quantifying generalization or empirically assessing important network features are difficult to apply to DNNs.

研究动机与目标

• 分析深度神经网络（DNNs）的经验成功与其行为缺乏理论理解之间的根本矛盾。
• 研究尽管DNNs具有高容量和过参数化，为何仍能良好泛化，这一现象违背了经典泛化理论。
• 探讨DNNs对对抗样本的敏感性，即微小的、难以察觉的扰动会导致巨大的预测误差。
• 考察优化动力学和训练算法在塑造DNN行为方面的作用，尤其是在高维非凸损失景观中。
• 评估当前理论框架是否能可靠地解释或预测DNN的性能、鲁棒性及失效模式。

提出的方法

• 综述现有的DNN理论方法，包括统计学习理论、函数逼近以及优化景观分析。
• 分析DNN与迭代优化算法（如基追踪去噪，BPDN）之间的联系，表明通过权重重用和非线性设计，DNN可实现近似算法。
• 研究非线性与权重重初始化在实现复杂函数逼近中的作用，与稀疏编码和压缩感知建立类比。
• 利用近似投影框架，形式化说明DNN如何通过分层迭代计算恢复正则化优化问题（如LASSO）的解。
• 考虑权重重用与残差连接在高效逼近迭代算法中的影响。
• 综述近期在理解DNN泛化方面的进展，包括信息瓶颈理论以及SGD中隐式正则化的作用。

实验结果

研究问题

• RQ1 为何深度神经网络尽管容量高且过参数化，仍能良好泛化？
• RQ2 DNN在多大程度上可被解释为实现已知优化算法（如BPDN或LASSO）？
• RQ3 DNN中的架构选择（如非线性、深度、权重重用）如何与稀疏编码或压缩感知问题的解相关联？
• RQ4 为何DNN对对抗样本如此脆弱，这揭示了其内部决策机制的哪些本质问题？
• RQ5 当前理论框架能否可靠地预测或解释DNN行为，包括鲁棒性与泛化能力？

主要发现

• DNN可被解释为实现迭代优化算法（如基追踪去噪，BPDN）的模型，其中每一层对应近似算法中的一个步骤。
• 通过使用近似算子，DNN与迭代求解器之间的等价性得到形式化，表明具有特定非线性和权重重用的DNN可求解正则化优化问题。
• 尽管存在这种形式化联系，DNN的泛化与鲁棒性在高维非凸设置下仍理解不足。
• 对抗样本揭示了DNN的根本脆弱性：微小且难以察觉的输入扰动可导致巨大的预测误差，挑战了其内部表征可靠性的观念。
• 信息瓶颈理论与SGD中的隐式正则化等理论框架提供了部分解释，但尚不足以完全理解DNN行为。
• 本文结论认为，DNN如同一种‘机器学习罗夏测验’——其可解释性在很大程度上取决于观察者的理论视角，凸显了建立更稳健、统一理论的迫切需求。