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 2104.02746cs.LGmath.FA
收藏

通过对神经网络近似空间的复杂性界进行采样来证明深度学习中的理论与实践差距

Proof of the Theory-to-Practice Gap in Deep Learning via Sampling Complexity bounds for Neural Network Approximation Spaces

University of Vienna(维也纳大学);Johann Radon Institute(约翰·拉东研究所)

Philipp Grohs, Felix Voigtlaender

摘要:

我们研究的计算复杂性的(确定性或随机)算法的基础上的点样本近似或集成的功能,可以很好地近似神经网络。这种算法(最突出的是随机梯度下降及其变体)在深度学习领域得到了广泛的应用。这一领域中最重要的问题之一是,是否有可能实现理论上可证明的神经网络逼近率的算法。我们回答这个问题的否定证明硬度结果的问题的近似和整合一类新的神经网络近似空间。特别是,我们的研究结果证实了深度学习中理论与实践之间的差距。我们补充我们的硬度的结果表明,近似率的一个可比的收敛顺序(至少在理论上)是可以实现的。

英文摘要:

We study the computational complexity of (deterministic or randomized) algorithms based on point samples for approximating or integrating functions that can be well approximated by neural networks. Such algorithms (most prominently stochastic gradient descent and its variants) are used extensively in the field of deep learning. One of the most important problems in this field concerns the question of whether it is possible to realize theoretically provable neural network approximation rates by such algorithms. We answer this question in the negative by proving hardness results for the problems of approximation and integration on a novel class of neural network approximation spaces. In particular, our results confirm a conjectured and empirically observed theory-to-practice gap in deep learning. We complement our hardness results by showing that approximation rates of a comparable order of convergence are (at least theoretically) achievable.

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。

↑