两篇新提交至arXiv的论文分析了深度神经网络中梯度下降方法的泛化性能。研究为使用GD和SGD训练的深度ReLU网络中的超额总体风险建立了minimax最优率，前提是网络宽度与深度和样本量成比例缩放。这些发现表明，具有足够宽度的深度神经网络可以实现与核方法相当的泛化率。 AI
影响 为深度学习泛化奠定了理论基础，可能指导未来的模型开发和分析。
排序理由 两篇发表在arXiv上的学术论文，详细介绍了深度学习泛化方面的理论进展。
AI 生成摘要 · Google Gemini · 来自 4 个来源。 我们如何撰写摘要 →
arXiv:2606.06764v1 Announce Type: new Abstract: Recent progress has been made in understanding the statistical generalization performance of gradient descent methods for overparameterized neural networks within the neural tangent kernel (NTK) regime. However, most of the existing…
arXiv:2606.06772v1 Announce Type: new Abstract: Understanding the generalization performance of over-parameterized neural networks has become a central topic in deep learning theory. While recent advances, particularly works under the Neural Tangent Kernel (NTK) regime, have shed…
Understanding the generalization performance of over-parameterized neural networks has become a central topic in deep learning theory. While recent advances, particularly works under the Neural Tangent Kernel (NTK) regime, have shed light on the behavior of shallow architectures,…
Recent progress has been made in understanding the statistical generalization performance of gradient descent methods for overparameterized neural networks within the neural tangent kernel (NTK) regime. However, most of the existing work on regression problems is limited to shall…