Abstract
In most practical settings and theoretical analysis, one assumes that a model can be trained until convergence. However, the growing complexity of machine learning datasets and models may violate such assumptions. Moreover, current approaches for hyper-parameter tuning and neural architecture search tend to be limited by practical resource constraints. Therefore, we introduce a formal setting for studying training under the non-asymptotic, resource-constrained regime, i.e. budgeted training. We analyze the following problem: "given a dataset, algorithm, and resource budget, what is the best achievable performance?" We focus on the number of optimization iterations as the representative resource. Under such a setting, we show that it is critical to adjust the learning rate schedule according to the given budget. Among budget-aware learning schedules, we find simple linear decay to be both robust and high-performing. We support our claim through extensive experiments with state-of-the-art models on ImageNet (image classification), Cityscapes (semantic segmentation), MS COCO (object detection and instance segmentation), and Kinetics (video classification). We also analyze our results and find that the key to a good schedule is budgeted convergence, a phenomenon whereby the gradient vanishes at the end of each allowed budget. We also revisit existing approaches for fast convergence, and show that budget-aware learning schedules readily outperform such approaches under (the practical but under-explored) budgeted setting.
Abstract (translated)
在大多数实际设置和理论分析中,假设模型可以训练到收敛。然而,机器学习数据集和模型的日益复杂可能会违反这些假设。此外,目前的超参数整定和神经架构搜索方法往往受到实际资源约束的限制。因此,我们引入了一个在非渐进的、资源受限的体制下学习培训的形式设置,即预算培训。我们分析以下问题:“给定一个数据集、算法和资源预算,什么是最佳的可实现性能?”我们把优化迭代的次数作为代表资源。在这种情况下,我们表明,根据给定的预算调整学习率计划是至关重要的。在了解预算的学习计划中,我们发现简单的线性衰减既健壮又高效。我们通过在Imagenet(图像分类)、Cityscapes(语义分割)、MS Coco(对象检测和实例分割)和Dynamics(视频分类)上的最新模型进行大量实验来支持我们的主张。我们还分析了我们的结果,发现一个好的时间表的关键是预算收敛,这是一种在每个允许的预算结束时梯度消失的现象。我们还回顾了现有的快速融合方法,并表明,预算感知学习计划在预算设置下(实际但探索不足)很容易优于这种方法。
URL
https://arxiv.org/abs/1905.04753