Mango: A Python Library for Parallel Hyperparameter Tuning
Mango: A Python Library for Parallel Hyperparameter Tuning
复制标题
Mango：用于并行超参数调优的 Python 库
DOI：
10.1109/icassp40776.2020.9054609
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
M. Srivastava
中科院分区：
文献类型：
--
作者：
S. Sandha;Mohit Aggarwal;Igor Fedorov;M. Srivastava
调整机器学习算法的超参数是一项繁琐的任务，通常是手动完成的。为了实现自动化的超参数调整，最近的工作已经开始使用基于贝叶斯优化的技术。然而，为了实际实现大规模机器学习训练管道的自动调整，现有库中仍然存在重大差距，包括缺乏抽象，容错和灵活性，以支持任何分布式计算框架上的调度。为了解决这些挑战，我们提出了Mango，一个用于并行超参数调优的Python库。Mango支持使用任何分布式调度框架，实现智能并行搜索策略，并提供丰富的抽象来定义与scikit-learn兼容的复杂超参数搜索空间。Mango在性能上与另一个广泛使用的库Hyperopt [1]相当。Mango是开源的[2]，目前在Arm Research的生产中使用，以提供最先进的超参数调整功能。
Tuning hyperparameters for machine learning algorithms is a tedious task, one that is typically done manually. To enable automated hyperparameter tuning, recent works have started to use techniques based on Bayesian optimization. However, to practically enable automated tuning for large scale machine learning training pipelines, significant gaps remain in existing libraries, including lack of abstractions, fault tolerance, and flexibility to support scheduling on any distributed computing framework. To address these challenges, we present Mango, a Python library for parallel hyperparameter tuning. Mango enables the use of any distributed scheduling framework, implements intelligent parallel search strategies, and provides rich abstractions for defining complex hyperparameter search spaces that are compatible with scikit-learn. Mango is comparable in performance to Hyperopt [1], another widely used library. Mango is available open-source [2] and is currently used in production at Arm Research to provide state-of-art hyperparameter tuning capabilities.
DOI：
10.17877/de290r-17800
发表时间：
2016
期刊：
--
影响因子：
--
作者：
Momchil Halstrup
通讯作者：
Momchil Halstrup