DeepLine: AutoML Tool for Pipelines Generation using Deep Reinforcement Learning and Hierarchical Actions Filtering

Created by W.Langdon from gp-bibliography.bib Revision:1.8098

@InProceedings{DBLP:conf/kdd/HeffetzVKR20,

• author = "Yuval Heffetz and Roman Vainshtein and Gilad Katz and Lior Rokach",
• title = "{DeepLine: AutoML} Tool for Pipelines Generation using Deep Reinforcement Learning and Hierarchical Actions Filtering",
• booktitle = "KDD 2020: The 26th ACM SIGKDD Conference on Knowledge Discovery and Data Mining",
• year = "2020",
• editor = "Rajesh Gupta and Yan Liu and Jiliang Tang and B. Aditya Prakash",
• pages = "2103--2113",
• address = "Virtual Event, CA, USA",
• month = aug # " 23-27",
• publisher = "ACM",
• keywords = "genetic algorithms, genetic programming, TPOT, AutoML, classification, deep reinforcement learning",
• timestamp = "Tue, 09 Mar 2021 09:46:47 +0100",
• biburl = "https://dblp.org/rec/conf/kdd/HeffetzVKR20.bib",
• bibsource = "dblp computer science bibliography, https://dblp.org",
• URL = "https://doi.org/10.1145/3394486.3403261",
• DOI = "doi:10.1145/3394486.3403261",
• size = "11 pages",
• abstract = "Automatic Machine Learning (AutoML) is an area of research aimed at automating Machine Learning (ML) activities that currently require the involvement of human experts. One of the most challenging tasks in this field is the automatic generation of end-to-end ML pipelines: combining multiple types of ML algorithms into a single architecture used for analysis of previously unseen data. This task has two challenging aspects: the first is the need to explore a large search space of algorithms and pipeline architectures. The second challenge is the computational cost of training and evaluating multiple pipelines. we present DeepLine, a reinforcement learning based approach for automatic pipeline generation. Our proposed approach uses an efficient representation of the search space together with a novel method for operating in environments with large and dynamic action spaces. By leveraging past knowledge gained from previously analysed datasets,our approach only needs to generate and evaluate few dozens of pipe lines to reach comparable or better performance than current state-of-the-art AutoML systems that evaluate hundreds and even thousands of pipelines in their optimisation process. Evaluation on 56 classification datasets demonstrates the merits of our approach",
• notes = "Comparison with TPOT and Auto-Sklearn etc",

}

Genetic Programming entries for Yuval Heffetz Roman Vainshtein Gilad Katz Lior Rokach

Citations