Pre-training (PT) followed by fine-tuning (FT) is an effective method for training neural networks, and has led to significant performance improvements in many domains. PT can incorporate various design choices such as task and data reweighting strategies, augmentation policies, and noise models, all of which can significantly impact the quality of representations learned. The hyperparameters introduced by these strategies therefore must be tuned appropriately. However, setting the values of these hyperparameters is challenging. Most existing methods either struggle to scale to high dimensions, are too slow and memory-intensive, or cannot be directly applied to the two-stage PT and FT learning process. In this work, we propose an efficient, gradient-based algorithm to meta-learn PT hyperparameters. We formalize the PT hyperparameter optimization problem and propose a novel method to obtain PT hyperparameter gradients by combining implicit differentiation and backpropagation through unrolled optimization. We demonstrate that our method improves predictive performance on two real-world domains. First, we optimize high-dimensional task weighting hyperparameters for multitask pre-training on protein-protein interaction graphs and improve AUROC by up to 3.9%. Second, we optimize a data augmentation neural network for self-supervised PT with SimCLR on electrocardiography data and improve AUROC by up to 1.9%.
翻译:培训前(PT)以及随后的微调(FT)是培训神经网络的有效方法,它导致许多领域显著的绩效改进。PT可以包括各种设计选择,如任务和数据加权战略、增强政策和噪音模型等,所有这些都可以对所学表现的质量产生重大影响。因此,这些战略引入的超参数必须适当调整。然而,设置这些超参数的数值具有挑战性。大多数现有方法要么是努力达到高度尺寸,过于缓慢,记忆密集,或者不能直接应用于两阶段PT和FT学习进程。在此工作中,我们提议一种高效的、基于梯度的算法,用于元的PT超参数超参数模型。我们将PT超参数优化问题正规化,并提议一种新颖的方法,以结合隐含的差别和无节制的优化来获得PT超参数梯度。我们的方法提高了两个现实世界领域的预测性能。首先,我们优化了高维度的超参数,用于多等值的PT和FT学习进程前PT。我们用S-S-Simalimal AS-toimalimalalal ASimalalalalalalal-toalalal 数据改进了ASlialalalalal-tomalial-toalalalalalalalalalalalalalalalalalalalmatoalalalalalmasal。我们Syal-Slimatoaltoal-Syal-Slial-S-Syal-Syal-Syal-Syal-Syalmatoalmatoal-Syal-Syal-Syal-Syal-Syal-Syal-Syal-Syal-Slimtoal-toal-SBLBBM 改进了第二个数据,我们SBAS的自我和SBBBSBS-S。我们S-SBSBBBBLBLBLBLBAS。我们通过S-S-S-S-S-S-S-SBS-S-S-S-S-S-S-S-S-S-S-S-S-SBSMASMASMASMASBSMASMASMASMASMASMA