用假设检验视角分析神经网络分类器，发现好泛化的模型会逐步逼近最优决策。
Implicit Hypothesis Testing and Divergence Preservation in Neural Network Representations
- 将分类任务视为类条件分布间的二元假设检验
- 训练中保留的KL散度持续增长，表明逼近奈曼-皮尔逊最优规则
- 提出证据-误差平面，可跨架构系统评估收敛性
我们从二元假设检验的角度研究神经分类器的训练动态。将分类重新形式化为由学习表征诱导的类条件分布之间的二元检验集合，并通过实验表明，在训练轨迹上，具有良好泛化能力的网络会逐渐趋近奈曼-皮尔逊最优决策规则，表现为学习表征所保留的KL散度单调增长。我们给出了达到精确最优的充分条件，讨论了其对训练正则化的影响，并定义了一个信息平面（称为证据-误差平面），可在不同网络架构间系统评估收敛性。
原文摘要 · Abstract (English)
We study the training dynamics of neural classifiers through the lens of binary hypothesis testing. We re-formalize classification as a collection of binary tests between class-conditional distributions induced by learned representations and show empirically that, along training trajectories, well-generalizing networks progressively approach Neyman-Pearson optimal decision rules, as measured by monotonic growth in the KL divergence retained by learned representations. We provide sufficient conditions for exact optimality, discuss its implications for training regularization, and define an informational plane, (so-called Evidence-Error plane) where convergence can be assessed methodically across network architecture.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。