AlphaD3M: Machine Learning Pipeline Synthesis

By Iddo Drori, Yamuna Krishnamurthy, Rémi Rampin, Raoni Lourenço, J. Ono, Kyunghyun Cho, Cláudio T. Silva, Juliana Freire. 2021

93 citations recorded by the source corpus.

We introduce AlphaD3M, an automatic machine learning (AutoML) system based on meta reinforcement learning using sequence models with self play. AlphaD3M is based on edit operations performed over machine learning pipeline primitives providing explainability. We compare AlphaD3M with state-of-the-art AutoML systems: Autosklearn, Autostacker, and TPOT, on OpenML datasets. AlphaD3M achieves competitive performance while being an order of magnitude faster, reducing computation time from hours to minutes, and is explainable by design.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Automated Machine Learning: State-of-The-Art and Open Challenges (2019, methodology): “AlphaD3M [33] has been introduced as an AutoML framework that uses meta reinforcement learning to find the most promising pipelines.”
Benchmark and Survey of Automated Machine Learning Frameworks (2019, methodology): “Pipeline structure search can also be considered as a game (Drori et al., 2018): an ML pipeline and the training data set represent the current board state s; for each step the player can choose between the three actions adding, removing or replacing a single node in the pipeline; the loss of the…”
Survey on Automated Machine Learning (2019, background): “Pipeline structure search can also be considered as a game (Drori et al., 2018): an ML pipeline and the training data set represent the current board state s; at each step the player can choose between the three actions a adding, removing or replacing a single element in the pipeline; the loss of…”
Meta-Learning: A Survey (2018, neutral): “AlphaD3M (Drori et al., 2018) uses a self-play reinforcement learning approach in which the current state is represented by the current pipeline, and actions include the addition, deletion, or replacement of pipeline components.”
Automatic Machine Learning by Pipeline Synthesis using Model-Based Reinforcement Learning and a Grammar (2019, methodology): “While AutoSklearn is limited to using the Sklearn library, AlphaD3M can be conﬁgured with machine learning primitives from other libraries.”
OpenML Benchmarking Suites (2017, background): “For example, Drori et al. [2018], Rakotoarison et al. [2019] and Gil et al. [2018], all published at the same workshop, each used different selections of datasets.”
Auto-sklearn: Efficient and Robust Automated Machine Learning (2019, methodology): “Since the publication of the original NIPS paper [14] Auto-sklearn has become a standard baseline for new approaches to automated machine learning such as FLASH [39], RECIPE [32], Hyperband [25], Auto-Prognosis [2], ML-PLAN [28], Auto-Stacker [10] and AlphaD3M [11].”
Data-centric Artificial Intelligence: A Survey (2023, unknown): “Basic manipulation Automation Learning-based [56] Augmentation data synthesis Automation Learning-based [79, 102, 104, 215] Upsampling Automation Programmatic [41, 95] Upsampling Automation Learning-based [272] Pipeline search Automation Pipeline [68, 76, 97, 132, 154, 159, 279]”