Using a thousand optimization tasks to learn hyperparameter search strategies
This work addresses the challenge of efficient hyperparameter tuning for machine learning practitioners, though it is incremental as it builds on existing meta-learning and optimization methods.
The authors tackled the problem of hyperparameter optimization by creating TaskSet, a large and diverse dataset of over a thousand tasks, and used it to meta-learn hyperparameter search strategies, achieving large speedups in sample efficiency over random search.
We present TaskSet, a dataset of tasks for use in training and evaluating optimizers. TaskSet is unique in its size and diversity, containing over a thousand tasks ranging from image classification with fully connected or convolutional neural networks, to variational autoencoders, to non-volume preserving flows on a variety of datasets. As an example application of such a dataset we explore meta-learning an ordered list of hyperparameters to try sequentially. By learning this hyperparameter list from data generated using TaskSet we achieve large speedups in sample efficiency over random search. Next we use the diversity of the TaskSet and our method for learning hyperparameter lists to empirically explore the generalization of these lists to new optimization tasks in a variety of settings including ImageNet classification with Resnet50 and LM1B language modeling with transformers. As part of this work we have opensourced code for all tasks, as well as ~29 million training curves for these problems and the corresponding hyperparameters.