Affiliation:
1. Graz University of Technology
Abstract
Data science pipelines are typically exploratory. An integral task of such pipelines are feature transformations, which transform raw data into numerical matrices or tensors for training or scoring. There exist a wide variety of transformations for different data modalities. These feature transformations incur large computational overhead due to expensive string processing and dictionary creation. Existing ML systems address this overhead by static parallelization schemes and interleaving transformations with model training. These approaches show good performance improvements for simple transformations, but struggle to handle different data characteristics (many features/distinct items) and multi-pass transformations. A key observation is that good parallelization strategies for feature transformations depend on data characteristics. In this paper, we introduce UPLIFT, a framework for
P
aralle
LI
zing
F
eature
T
ransformations. UPLIFT constructs a fine-grained task graph for a set of transformations, optimizes the plan according to data characteristics, and executes this plan in a cache-conscious manner. We show that the resulting framework is applicable to a wide range of transformations. Furthermore, we propose the FTBench benchmark with transformations and datasets from various domains. On this benchmark, UPLIFT yields speedups of up to 31.6x (9.27x on average) compared to state-of-the-art ML systems.
Publisher
Association for Computing Machinery (ACM)
Subject
General Earth and Planetary Sciences,Water Science and Technology,Geography, Planning and Development
Reference120 articles.
1. 2021. Joblib: running Python functions as pipeline jobs. https://joblib.readthedocs.io/en/latest/ Accessed: 2022-07-11. 2021. Joblib: running Python functions as pipeline jobs. https://joblib.readthedocs.io/en/latest/ Accessed: 2022-07-11.
2. 2022. Citation Network Dataset. https://www.aminer.org/citation Accessed: 2022-07-11. 2022. Citation Network Dataset. https://www.aminer.org/citation Accessed: 2022-07-11.
3. Martín Abadi et al. 2016. TensorFlow: A System for Large-Scale Machine Learning. In OSDI. 265--283. Martín Abadi et al. 2016. TensorFlow: A System for Large-Scale Machine Learning. In OSDI. 265--283.
4. Mind the gap
5. Dakshi Agrawal et al. 2015. SparkBench - A Spark Performance Testing Suite. In TPCTC@VLDB Workshop. Dakshi Agrawal et al. 2015. SparkBench - A Spark Performance Testing Suite. In TPCTC@VLDB Workshop.
Cited by
4 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献