arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1706.04769stat.MLcs.LG
收藏

基于逐次凸逼近的神经网络随机训练

Stochastic Training of Neural Networks via Successive Convex Approximations

Sapienza University of Rome(罗马大学);University of Perugia(佩鲁贾大学)

Simone Scardapane, Paolo Di Lorenzo

摘要:

提出了一类新的神经网络训练算法。这些都是基于非凸优化领域的最新发展,在连续凸逼近(SCA)技术的一般名称。其基本思想是迭代地用一系列(强凸)近似值替换原始(非凸,高维)学习问题,这些近似值既精确又易于优化。从类似的想法(例如,准牛顿算法),可以仅使用神经网络函数的一阶信息以随机方式构造逼近,同时利用学习问题的整体结构以实现更快的收敛。我们讨论了几个用例,基于损失函数的不同选择(例如,平方损失和交叉熵损失),以及NN权重的正则化。我们在几个中等规模的基准问题上进行了实验,并在涉及模拟物理数据的大规模数据集上进行了实验。结果表明,该算法如何优于最先进的技术,提供更快的收敛到一个更好的最小值。此外,我们展示了如何可以很容易地在多个计算单元并行化的算法,而不妨碍其性能。特别地,每个计算单元可以优化在输入变量的随机分配的子集上定义的定制的代理函数,其维度可以完全取决于可用的计算能力来选择。

英文摘要:

This paper proposes a new family of algorithms for training neural networks (NNs). These are based on recent developments in the field of non-convex optimization, going under the general name of successive convex approximation (SCA) techniques. The basic idea is to iteratively replace the original (non-convex, highly dimensional) learning problem with a sequence of (strongly convex) approximations, which are both accurate and simple to optimize. Differently from similar ideas (e.g., quasi-Newton algorithms), the approximations can be constructed using only first-order information of the neural network function, in a stochastic fashion, while exploiting the overall structure of the learning problem for a faster convergence. We discuss several use cases, based on different choices for the loss function (e.g., squared loss and cross-entropy loss), and for the regularization of the NN's weights. We experiment on several medium-sized benchmark problems, and on a large-scale dataset involving simulated physical data. The results show how the algorithm outperforms state-of-the-art techniques, providing faster convergence to a better minimum. Additionally, we show how the algorithm can be easily parallelized over multiple computational units without hindering its performance. In particular, each computational unit can optimize a tailored surrogate function defined on a randomly assigned subset of the input variables, whose dimension can be selected depending entirely on the available computational power.

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。

↑