pith. sign in

arxiv: 2310.02540 · v2 · submitted 2023-10-04 · 💻 cs.LG · cs.AI· cs.DB· cs.IR

Auto-FP: An Experimental Study of Automated Feature Preprocessing for Tabular Data

Danrui Qi , Jinglin Peng , Yongjun He , Jiannan Wang This is my paper

Pith reviewed 2026-05-24 05:56 UTC · model grok-4.3

classification 💻 cs.LG cs.AIcs.DBcs.IR
keywords automated feature preprocessingtabular datahyperparameter optimizationneural architecture searchAutoMLfeature engineeringmachine learningsearch algorithms
0
0 comments X

The pith

Auto-FP can be solved by modeling it as hyperparameter optimization or neural architecture search.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that the problem of automatically selecting and ordering feature preprocessors for tabular data can be addressed by recasting it as a hyperparameter optimization or neural architecture search task. This modeling allows the direct application of existing HPO and NAS algorithms to the Auto-FP problem. Through experiments with 15 algorithms on 45 public datasets, evolution-based methods show the best average performance while random search proves surprisingly competitive. Advanced methods like surrogate-model-based and bandit-based searches often do not improve upon random search in this domain. The study also explores parameter search extensions and limitations in current AutoML tools.

Core claim

Auto-FP can be modelled as either a hyperparameter optimization (HPO) or a neural architecture search (NAS) problem, which enables extending a variety of HPO and NAS algorithms to solve the Auto-FP problem.

What carries the argument

The modeling of automated feature preprocessing as an HPO or NAS problem to enable reuse of existing search algorithms.

If this is right

  • Evolution-based algorithms achieve the leading average ranking among the tested methods.
  • Random search is a strong baseline that outperforms many surrogate-model-based and bandit-based algorithms for Auto-FP.
  • Analysis reveals reasons why standard HPO and NAS methods underperform random search in this setting.
  • Auto-FP can be extended to support parameter search using two different approaches.
  • Popular AutoML tools have limitations in handling automated feature preprocessing.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Improvements to HPO and NAS algorithms could automatically translate to better Auto-FP performance.
  • The bottleneck analysis suggests specific areas where new Auto-FP tailored algorithms could be developed.
  • Adopting this approach in practice could significantly reduce the manual effort in building ML pipelines for tabular data.
  • Future work might test these methods on private industry datasets to confirm generalizability.

Load-bearing premise

The large combinatorial search space of preprocessor choices and orderings can be effectively navigated by standard HPO and NAS algorithms without requiring domain-specific adaptations or becoming computationally intractable.

What would settle it

If applying HPO and NAS algorithms to Auto-FP fails to produce preprocessing pipelines that improve model quality over default or manual choices on most of the 45 datasets, the modeling approach would be falsified.

Figures

Figures reproduced from arXiv: 2310.02540 by Danrui Qi, Jiannan Wang, Jinglin Peng, Yongjun He.

Figure 1
Figure 1. Figure 1: Illustration of different feature preprocessors. (a) No [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Distribution of LR accuracy with different feature preprocessing pipelines. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The analogy between HPO and Auto-FP. The automated feature preprocessing (Auto-FP) problem aims to find the best pipeline with minimal error, i.e., arg min L ∈S𝑝𝑖𝑝𝑒 error(L) 3.2 Auto-FP as HPO and NAS Interestingly, Auto-FP can be viewed in two different ways. Auto-FP as HPO. HPO aims to find the best combination of clas￾sifier and its related hyperparameters for a given dataset. Since different classifier… view at source ↗
Figure 4
Figure 4. Figure 4: The anology between NAS and Auto-FP. is that it puts a feature preprocessor (e.g., Normalizer) rather than a neural-architecture operator at each position. Remark. HPO, NAS and AutoFP tasks all aim to find the best combination in a large search space. HPO aims to find the best combination of the classifier and its related hyperparameters for a given dataset. NAS aims to find the best combination of operato… view at source ↗
Figure 5
Figure 5. Figure 5: Statistics of 45 real-world ML datasets. [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Parameter adjustment for Hyperband and BOHB. The upper two vary 𝜂. The lower two vary min_budget. Even with several parameter adjustments, Hyperband and BOHB still cannot outperform RS.    +)*( #! % &( , )*+(  ! )  # %  !%   #!%     +)*( #! % &( , )*+(  ! )  # %  !%   #!%      [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Overhead Percentage on 7 datasets with different downstream ML models. “Pick” means the overhead of picking up [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Comparison of One-step and Two-step in the extended low-cardinality search space in Table 6. One-step is preferred. cardinality is the cardinality of n_quantiles, which is 1990. [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Comparison of One-step and Two-step in the extended high-cardinality search space in Table 7. Two-step is preferred.   " [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗
Figure 11
Figure 11. Figure 11: Evaluate Auto-FP in an AutoML context (extended [PITH_FULL_IMAGE:figures/full_fig_p014_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: The gap of the validation accuracy between random search and other search algorithms on all datasets with 60s and [PITH_FULL_IMAGE:figures/full_fig_p024_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: The gap of the validation accuracy between random search and other search algorithms on all datasets with 60s and [PITH_FULL_IMAGE:figures/full_fig_p025_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: The gap of the validation accuracy between random search and other search algorithms on all datasets with 60s and [PITH_FULL_IMAGE:figures/full_fig_p026_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: The gap of the validation accuracy between random search and other search algorithms on all datasets with 60s and [PITH_FULL_IMAGE:figures/full_fig_p027_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: The gap of the validation accuracy between random search and other search algorithms on all datasets with 60s and [PITH_FULL_IMAGE:figures/full_fig_p028_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: The trend of validation accuracy with the increase of time limit on all datasets - part 1. [PITH_FULL_IMAGE:figures/full_fig_p029_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: The trend of validation accuracy with the increase of time limit on all datasets - part 2. [PITH_FULL_IMAGE:figures/full_fig_p030_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: The trend of validation accuracy with the increase of time limit on all datasets - part 3. [PITH_FULL_IMAGE:figures/full_fig_p031_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: Overhead percentage of all datasets with different downstream ML models. “Pic” means the overhead of picking up [PITH_FULL_IMAGE:figures/full_fig_p032_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: Overhead percentage of all datasets with different downstream ML models. “Pic” means the overhead of picking up [PITH_FULL_IMAGE:figures/full_fig_p033_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: Overhead percentage of all datasets with different downstream ML models. “Pic” means the overhead of picking up [PITH_FULL_IMAGE:figures/full_fig_p034_22.png] view at source ↗
Figure 23
Figure 23. Figure 23: Comparison of One-step and Two-step in the extended low-cardinality search space in Table 5 (related to Figure 8) - [PITH_FULL_IMAGE:figures/full_fig_p035_23.png] view at source ↗
Figure 24
Figure 24. Figure 24: Comparison of One-step and Two-step in the extended low-cardinality search space in Table 5 (related to Figure 8) - [PITH_FULL_IMAGE:figures/full_fig_p036_24.png] view at source ↗
Figure 25
Figure 25. Figure 25: Comparison of One-step and Two-step in the extended low-cardinality search space in Table 5 (related to Figure 8) - [PITH_FULL_IMAGE:figures/full_fig_p037_25.png] view at source ↗
Figure 26
Figure 26. Figure 26: Comparison of One-step and Two-step in the extended low-cardinality search space in Table 6 (related to Figure 9) - [PITH_FULL_IMAGE:figures/full_fig_p038_26.png] view at source ↗
Figure 27
Figure 27. Figure 27: Comparison of One-step and Two-step in the extended low-cardinality search space in Table 6 (related to Figure 9) - [PITH_FULL_IMAGE:figures/full_fig_p039_27.png] view at source ↗
Figure 28
Figure 28. Figure 28: Comparison of One-step and Two-step in the extended low-cardinality search space in Table 6 (related to Figure 9) - [PITH_FULL_IMAGE:figures/full_fig_p040_28.png] view at source ↗
Figure 29
Figure 29. Figure 29: Evaluate Auto-FP in an AutoML context (default search space, related to Figure 10). [PITH_FULL_IMAGE:figures/full_fig_p041_29.png] view at source ↗
Figure 30
Figure 30. Figure 30: Evaluate Auto-FP in an AutoML context (extended search space in Table 5, related to Figure 11). [PITH_FULL_IMAGE:figures/full_fig_p042_30.png] view at source ↗
read the original abstract

Classical machine learning models, such as linear models and tree-based models, are widely used in industry. These models are sensitive to data distribution, thus feature preprocessing, which transforms features from one distribution to another, is a crucial step to ensure good model quality. Manually constructing a feature preprocessing pipeline is challenging because data scientists need to make difficult decisions about which preprocessors to select and in which order to compose them. In this paper, we study how to automate feature preprocessing (Auto-FP) for tabular data. Due to the large search space, a brute-force solution is prohibitively expensive. To address this challenge, we interestingly observe that Auto-FP can be modelled as either a hyperparameter optimization (HPO) or a neural architecture search (NAS) problem. This observation enables us to extend a variety of HPO and NAS algorithms to solve the Auto-FP problem. We conduct a comprehensive evaluation and analysis of 15 algorithms on 45 public ML datasets. Overall, evolution-based algorithms show the leading average ranking. Surprisingly, the random search turns out to be a strong baseline. Many surrogate-model-based and bandit-based search algorithms, which achieve good performance for HPO and NAS, do not outperform random search for Auto-FP. We analyze the reasons for our findings and conduct a bottleneck analysis to identify the opportunities to improve these algorithms. Furthermore, we explore how to extend Auto-FP to support parameter search and compare two ways to achieve this goal. In the end, we evaluate Auto-FP in an AutoML context and discuss the limitations of popular AutoML tools. To the best of our knowledge, this is the first study on automated feature preprocessing. We hope our work can inspire researchers to develop new algorithms tailored for Auto-FP.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper claims that automated feature preprocessing (Auto-FP) for tabular data can be modeled as either a hyperparameter optimization (HPO) or neural architecture search (NAS) problem. This modeling enables extending a variety of existing HPO and NAS algorithms to search over preprocessor selections and orderings. A comprehensive evaluation of 15 algorithms across 45 public ML datasets finds that evolution-based methods achieve the leading average rankings, random search is a strong baseline, and many surrogate-model-based and bandit-based methods fail to outperform random search. The work includes analysis of these findings, a bottleneck analysis, an extension to joint parameter search, and an evaluation of Auto-FP within an AutoML context, presenting itself as the first dedicated study on the topic.

Significance. If the central modeling claim and empirical rankings hold after verification of experimental details, the paper provides a useful empirical foundation for Auto-FP by showing that standard HPO/NAS methods can be applied but that surrogate and bandit approaches do not transfer well, thereby identifying a need for domain-tailored algorithms. The scale of the evaluation (45 datasets) and the explicit bottleneck analysis are strengths that could guide future work. The observation that random search remains competitive is a falsifiable insight worth documenting.

major comments (3)
  1. [Abstract / implied Section 3] Abstract / implied modeling in Section 3: the claim that preprocessor selection and ordering can be directly encoded as a fixed-dimensional HPO problem or NAS cell without domain-specific adaptations is load-bearing for the central contribution. The reported result that surrogate-model and bandit methods fail to beat random search is consistent with the possibility that variable-length sequences and inter-preprocessor interactions are not adequately captured by standard encodings, undermining the assertion that existing HPO/NAS algorithms can be extended without further machinery.
  2. [Evaluation on 45 datasets] Evaluation section (45 datasets): the soundness of the average-ranking claims depends on verifiable details of data splits, cross-validation procedure, and statistical testing for ranking differences. Without these, it is impossible to rule out that variance, post-hoc dataset selection, or improper multiple-testing correction affects the conclusion that evolution-based algorithms lead while others do not.
  3. [Bottleneck analysis] Bottleneck analysis: the analysis should explicitly test whether the observed lack of structure for surrogate/bandit methods originates from the HPO/NAS modeling choice itself (e.g., loss of ordering semantics) rather than solely from algorithmic limitations; otherwise the recommendation to develop new algorithms for Auto-FP rests on an unexamined premise.
minor comments (2)
  1. [Abstract] The abstract states that 'we analyze the reasons for our findings' but does not point to the specific section; adding an explicit cross-reference would improve readability.
  2. [Modeling section] Notation for how preprocessor pipelines are encoded as fixed-length vectors or architecture cells should be clarified with a small example in the modeling section to make the HPO/NAS reduction concrete.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major comment below and indicate planned revisions.

read point-by-point responses

  1. Referee: [Abstract / implied Section 3] Abstract / implied modeling in Section 3: the claim that preprocessor selection and ordering can be directly encoded as a fixed-dimensional HPO problem or NAS cell without domain-specific adaptations is load-bearing for the central contribution. The reported result that surrogate-model and bandit methods fail to beat random search is consistent with the possibility that variable-length sequences and inter-preprocessor interactions are not adequately captured by standard encodings, undermining the assertion that existing HPO/NAS algorithms can be extended without further machinery.

    Authors: We maintain that modeling Auto-FP as HPO or NAS is valid because it permits direct extension of existing algorithms, as specified in Section 3 via concrete encodings (fixed-dimensional for HPO; cell-based for NAS). These encodings approximate variable-length sequences and interactions through padding and ordering constraints, but they require no new algorithmic machinery. The underperformance of surrogate/bandit methods is reported as an empirical observation about the resulting search space, not as evidence against the modeling itself. We will revise the abstract and Section 3 to state the approximations more explicitly and note their possible effect on algorithm transfer. revision: partial

  2. Referee: [Evaluation on 45 datasets] Evaluation section (45 datasets): the soundness of the average-ranking claims depends on verifiable details of data splits, cross-validation procedure, and statistical testing for ranking differences. Without these, it is impossible to rule out that variance, post-hoc dataset selection, or improper multiple-testing correction affects the conclusion that evolution-based algorithms lead while others do not.

    Authors: The Evaluation section already specifies the use of standard train/test splits from the source repositories, 5-fold cross-validation, average ranks, and Friedman/Nemenyi tests. To improve verifiability we will expand the section with explicit statements on dataset selection criteria, exact multiple-testing procedure, and variance handling. We will also release the full experimental code and splits. revision: yes

  3. Referee: [Bottleneck analysis] Bottleneck analysis: the analysis should explicitly test whether the observed lack of structure for surrogate/bandit methods originates from the HPO/NAS modeling choice itself (e.g., loss of ordering semantics) rather than solely from algorithmic limitations; otherwise the recommendation to develop new algorithms for Auto-FP rests on an unexamined premise.

    Authors: The bottleneck analysis examines empirical performance gaps and attributes them in part to preprocessor interaction complexity. We did not run an explicit ablation that isolates encoding effects from algorithmic limitations. We will add a paragraph discussing how the chosen encodings may contribute to the observed lack of exploitable structure, while clarifying that the recommendation for new algorithms is based on the overall empirical pattern rather than a single causal claim. revision: partial

Circularity Check

0 steps flagged

No circularity: pure empirical benchmarking with external validation

full rationale

The paper is an experimental study that models Auto-FP as HPO/NAS by observation and then benchmarks 15 algorithms on 45 public datasets. No equations, fitted parameters, or first-principles derivations are present that could reduce to inputs by construction. Rankings and conclusions are measured against external public data; the modeling step is a framing choice, not a self-referential derivation. No self-citation load-bearing steps or ansatz smuggling occur. This matches the default expectation of a non-circular empirical paper.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an empirical experimental paper; it introduces no mathematical derivations, fitted constants, or new postulated entities. The only implicit assumptions are standard ones about the validity of public benchmark datasets and the transferability of HPO/NAS algorithms to the preprocessing domain.

pith-pipeline@v0.9.0 · 5865 in / 1098 out tokens · 26473 ms · 2026-05-24T05:56:09.008140+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. FELA: A Multi-Agent Evolutionary System for Feature Engineering of Industrial Event Log Data

    cs.AI 2025-10 unverdicted novelty 6.0

    FELA deploys specialized LLM agents in an evolutionary framework to generate, validate, and refine explainable features from heterogeneous industrial event logs, improving downstream model performance.

Reference graph

Works this paper leans on

98 extracted references · 98 canonical work pages · cited by 1 Pith paper · 8 internal anchors

  1. [1]

    AutoML Challenge Website

    2021. AutoML Challenge Website. https://automl.chalearn.org/data

    work page 2021

  2. [2]

    Entropy Function in Scipy Library

    2021. Entropy Function in Scipy Library. https://docs.scipy.org/doc/scipy/ reference/generated/scipy.stats.entropy.html

    work page 2021

  3. [3]

    HpBandSter

    2021. HpBandSter. https://automl.github.io/HpBandSter/build/html/index.html

    work page 2021

  4. [4]

    Kaggle Datasets

    2021. Kaggle Datasets. https://www.kaggle.com/datasets. 14

    work page 2021

  5. [5]

    Kurtosis Function in Scipy Library

    2021. Kurtosis Function in Scipy Library. https://docs.scipy.org/doc/scipy/ reference/generated/scipy.stats.kurtosis.html

    work page 2021

  6. [6]

    Scikit-learn: Machine Learning in Python

    2021. Scikit-learn: Machine Learning in Python. https://scikit-learn.org/stable/

    work page 2021

  7. [7]

    Skewness Function in Scipy Library

    2021. Skewness Function in Scipy Library. https://docs.scipy.org/doc/scipy/ reference/generated/scipy.stats.skew.html

    work page 2021

  8. [8]

    State of Data Science and Machine Learning 2021

    2021. State of Data Science and Machine Learning 2021. https://www.kaggle. com/kaggle-survey-2021

    work page 2021

  9. [9]

    Auto-FP: An Experimental Study of Automated Feature Preprocessing for Tabular Data (Technical Report)

    2022. Auto-FP: An Experimental Study of Automated Feature Preprocessing for Tabular Data (Technical Report). (2022). https://github.com/AutoFP/Auto- FP/blob/main/Auto-FP(technical_report).pdf

    work page 2022

  10. [10]

    Scikit-learn Documentation

    2022. Scikit-learn Documentation. https://scikit-learn.org/stable/modules/ preprocessing.html

    work page 2022

  11. [11]

    Scikit-learn: Preprocessing Data

    2022. Scikit-learn: Preprocessing Data. https://scikit-learn.org/stable/modules/ preprocessing.html

    work page 2022

  12. [12]

    Peter Auer, Nicolo Cesa-Bianchi, and Paul Fischer. 2002. Finite-time analysis of the multiarmed bandit problem. Machine learning 47 (2002), 235–256

    work page 2002

  13. [13]

    Wolfgang Banzhaf, Peter Nordin, Robert E Keller, and Frank D Francone. 1998. Genetic programming: an introduction: on the automatic evolution of computer programs and its applications . Morgan Kaufmann Publishers Inc

    work page 1998

  14. [14]

    Rémi Bardenet, Mátyás Brendel, Balázs Kégl, and Michèle Sebag. 2013. Col- laborative hyperparameter tuning. In Proceedings of the 30th International Conference on Machine Learning, ICML 2013, Atlanta, GA, USA, 16-21 June 2013 (JMLR Workshop and Conference Proceedings) , Vol. 28. JMLR.org, 199–207. http://proceedings.mlr.press/v28/bardenet13.html

    work page 2013

  15. [15]

    James Bergstra, Rémi Bardenet, Yoshua Bengio, and Balázs Kégl. 2011. Algorithms for Hyper-Parameter Optimization. In Advances in Neural Information Processing Systems 24: 25th Annual Conference on Neural Information Processing Systems

    work page 2011

  16. [16]

    Zemel, Peter L

    Proceedings of a meeting held 12-14 December 2011, Granada, Spain , John Shawe-Taylor, Richard S. Zemel, Peter L. Bartlett, Fernando C. N. Pereira, and Kilian Q. Weinberger (Eds.). 2546–2554. https://proceedings.neurips.cc/paper/ 2011/hash/86e8f7ab32cfd12577bc2619bc635690-Abstract.html

    work page 2011

  17. [17]

    James Bergstra and Yoshua Bengio. 2012. Random Search for Hyper-Parameter Optimization. Journal of Machine Learning Research 13, 10 (2012), 281–305. http://jmlr.org/papers/v13/bergstra12a.html

    work page 2012

  18. [18]

    James Bergstra, Dan Yamins, David D Cox, et al . 2013. Hyperopt: A python library for optimizing the hyperparameters of machine learning algorithms. In Proceedings of the 12th Python in science conference , Vol. 13. Citeseer, 20

    work page 2013

  19. [19]

    Bernhard E Boser, Isabelle M Guyon, and Vladimir N Vapnik. 1992. A train- ing algorithm for optimal margin classifiers. In Proceedings of the fifth annual workshop on Computational learning theory . 144–152

    work page 1992

  20. [20]

    Jason Brownlee. 2020. Data preparation for machine learning: data cleaning, feature selection, and data transforms in Python

    work page 2020

  21. [21]

    Girish Chandrashekar and Ferat Sahin. 2014. A survey on feature selection methods. Computers & Electrical Engineering 40, 1 (2014), 16–28

    work page 2014

  22. [22]

    Krishna Teja Chitty-Venkata, Murali Emani, Venkatram Vishwanath, and Arun K. Somani. 2023. Neural Architecture Search Benchmarks: Insights and Survey.IEEE Access 11 (2023), 25217–25236. https://doi.org/10.1109/ACCESS.2023.3253818

    work page doi:10.1109/access.2023.3253818 2023

  23. [23]

    Xuanyi Dong and Yi Yang. 2020. NAS-Bench-201: Extending the Scope of Repro- ducible Neural Architecture Search. In 8th International Conference on Learning Representations, ICLR 2020, Addis Ababa, Ethiopia, April 26-30, 2020 . OpenRe- view.net. https://openreview.net/forum?id=HJxyZkBKDr

    work page 2020

  24. [24]

    Stefan Falkner, Aaron Klein, and Frank Hutter. 2018. BOHB: Robust and Ef- ficient Hyperparameter Optimization at Scale. In Proceedings of the 35th In- ternational Conference on Machine Learning, ICML 2018, Stockholmsmässan, Stockholm, Sweden, July 10-15, 2018 (Proceedings of Machine Learning Research) , Jennifer G. Dy and Andreas Krause (Eds.), Vol. 80. P...

    work page 2018

  25. [25]

    Matthias Feurer and Frank Hutter. 2019. Hyperparameter optimization. In Automated machine learning. Springer, Cham, 3–33

    work page 2019

  26. [26]

    Matthias Feurer, Aaron Klein, Katharina Eggensperger, Jost Tobias Springenberg, Manuel Blum, and Frank Hutter. 2015. Efficient and Robust Automated Machine Learning. In Proceedings of the 28th International Conference on Neural Information Processing Systems - Volume 2 (NIPS’15). 2755–2763

    work page 2015

  27. [27]

    George Forman et al. 2003. An extensive empirical study of feature selection metrics for text classification. J. Mach. Learn. Res. 3, Mar (2003), 1289–1305

    work page 2003

  28. [28]

    An Open Source AutoML Benchmark

    P. Gijsbers, E. LeDell, S. Poirier, J. Thomas, B. Bischl, and J. Vanschoren. 2019. An Open Source AutoML Benchmark. arXiv preprint arXiv:1907.00909 [cs.LG] (2019). https://arxiv.org/abs/1907.00909 Accepted at AutoML Workshop at ICML 2019

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  29. [29]

    Daniel Golovin, Benjamin Solnik, Subhodeep Moitra, Greg Kochanski, John Karro, and D. Sculley. 2017. Google Vizier: A Service for Black-Box Optimization. In Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Halifax, NS, Canada, August 13 - 17, 2017 . ACM, 1487–1495. https://doi.org/10.1145/3097983.3098043

    work page doi:10.1145/3097983.3098043 2017

  30. [30]

    Gongde Guo, Hui Wang, David Bell, Yaxin Bi, and Kieran Greer. 2003. KNN model-based approach in classification. In On The Move to Meaningful Internet Systems 2003: CoopIS, DOA, and ODBASE: OTM Confederated International Con- ferences, CoopIS, DOA, and ODBASE 2003, Catania, Sicily, Italy, November 3-7,

    work page 2003

  31. [31]

    Springer, 986–996

    Proceedings. Springer, 986–996

    work page

  32. [32]

    Huifeng Guo, Ruiming Tang, Yunming Ye, Zhenguo Li, and Xiuqiang He. 2017. DeepFM: A Factorization-Machine based Neural Network for CTR Prediction. In Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence, IJCAI 2017, Melbourne, Australia, August 19-25, 2017 , Carles Sierra (Ed.). ijcai.org, 1725–1731. https://doi.org/10...

    work page doi:10.24963/ijcai.2017/239 2017

  33. [33]

    PC Hammer. 1962. Adaptive control processes: a guided tour (R. Bellman)

    work page 1962

  34. [34]

    Jiawei Han, Jian Pei, and Yiwen Yin. 2000. Mining Frequent Patterns without Candidate Generation. In Proceedings of the 2000 ACM SIGMOD International Conference on Management of Data, May 16-18, 2000, Dallas, Texas, USA , Weidong Chen, Jeffrey F. Naughton, and Philip A. Bernstein (Eds.). ACM, 1–12. https: //doi.org/10.1145/342009.335372

    work page doi:10.1145/342009.335372 2000

  35. [35]

    Xin He, Kaiyong Zhao, and Xiaowen Chu. 2021. AutoML: A survey of the state- of-the-art. Knowl. Based Syst. 212 (2021), 106622. https://doi.org/10.1016/j.knosys. 2020.106622

    work page doi:10.1016/j.knosys 2021

  36. [36]

    Franziska Horn, Robert Pack, and Michael Rieger. 2019. The autofeat Python Li- brary for Automated Feature Engineering and Selection. InMachine Learning and Knowledge Discovery in Databases - International Workshops of ECML PKDD 2019, Würzburg, Germany, September 16-20, 2019, Proceedings, Part I (Communications in Computer and Information Science) , Peggy...

    work page doi:10.1007/978-3-030-43823-4_10 2019

  37. [37]

    Frank Hutter, Holger H Hoos, and Kevin Leyton-Brown. 2011. Sequential model- based optimization for general algorithm configuration. In International confer- ence on learning and intelligent optimization . Springer, 507–523

    work page 2011

  38. [38]

    Frank Hutter, Holger H Hoos, Kevin Leyton-Brown, and Kevin Murphy. 2010. Time-bounded sequential parameter optimization. In International Conference on Learning and Intelligent Optimization . Springer, 281–298

    work page 2010

  39. [39]

    Population Based Training of Neural Networks

    Max Jaderberg, Valentin Dalibard, Simon Osindero, Wojciech M. Czarnecki, Jeff Donahue, Ali Razavi, Oriol Vinyals, Tim Green, Iain Dunning, Karen Simonyan, Chrisantha Fernando, and Koray Kavukcuoglu. 2017. Population Based Training of Neural Networks. CoRR abs/1711.09846 (2017). arXiv:1711.09846 http://arxiv. org/abs/1711.09846

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  40. [40]

    Gareth James, Daniela Witten, Trevor Hastie, and Robert Tibshirani. 2013. An introduction to statistical learning . Vol. 112. Springer

    work page 2013

  41. [41]

    Jamieson and Ameet Talwalkar

    Kevin G. Jamieson and Ameet Talwalkar. 2016. Non-stochastic Best Arm Identification and Hyperparameter Optimization. In Proceedings of the 19th International Conference on Artificial Intelligence and Statistics, AISTATS 2016, Cadiz, Spain, May 9-11, 2016 (JMLR Workshop and Conference Proceedings) , Arthur Gretton and Christian C. Robert (Eds.), Vol. 51. J...

    work page 2016

  42. [42]

    Ian T Jolliffe and Jorge Cadima. 2016. Principal component analysis: a review and recent developments. Philosophical transactions of the royal society A: Math- ematical, Physical and Engineering Sciences 374, 2065 (2016), 20150202

    work page 2016

  43. [43]

    Gilad Katz, Eui Chul Richard Shin, and Dawn Song. 2016. ExploreKit: Automatic Feature Generation and Selection. In IEEE 16th International Conference on Data Mining, ICDM 2016, December 12-15, 2016, Barcelona, Spain , Francesco Bonchi, Josep Domingo-Ferrer, Ricardo Baeza-Yates, Zhi-Hua Zhou, and Xindong Wu (Eds.). IEEE Computer Society, 979–984. https://d...

    work page doi:10.1109/icdm.2016.0123 2016

  44. [44]

    Scott Kirkpatrick, C Daniel Gelatt, and Mario P Vecchi. 1983. Optimization by simulated annealing. science 220, 4598 (1983), 671–680

    work page 1983

  45. [45]

    Pang Wei Koh and Percy Liang. 2017. Understanding Black-box Predictions via Influence Functions. InProceedings of the 34th International Conference on Machine Learning, ICML 2017, Sydney, NSW, Australia, 6-11 August 2017 (Proceedings of Machine Learning Research) , Doina Precup and Yee Whye Teh (Eds.), Vol. 70. PMLR, 1885–1894. http://proceedings.mlr.pres...

    work page 2017

  46. [46]

    Sanjay Krishnan and Eugene Wu. 2019. AlphaClean: Automatic Generation of Data Cleaning Pipelines. CoRR abs/1904.11827 (2019). arXiv:1904.11827 http://arxiv.org/abs/1904.11827

    work page internal anchor Pith review Pith/arXiv arXiv 2019

  47. [47]

    Max Kuhn and Kjell Johnson. 2021. Feature engineering and selection: A practical approach for predictive models . Chapman & Hall/CRC Press

    work page 2021

  48. [48]

    Doris Jung Lin Lee, Stephen Macke, Doris Xin, Angela Lee, Silu Huang, and Aditya G Parameswaran. 2019. A Human-in-the-loop Perspective on AutoML: Milestones and the Road Ahead. IEEE Data Eng. Bull. 42, 2 (2019), 59–70

    work page 2019

  49. [49]

    Jundong Li, Kewei Cheng, Suhang Wang, Fred Morstatter, Robert P Trevino, Jiliang Tang, and Huan Liu. 2017. Feature selection: A data perspective. ACM computing surveys (CSUR) 50, 6 (2017), 1–45

    work page 2017

  50. [50]

    Jamieson, Giulia DeSalvo, Afshin Rostamizadeh, and Ameet Talwalkar

    Lisha Li, Kevin G. Jamieson, Giulia DeSalvo, Afshin Rostamizadeh, and Ameet Talwalkar. 2017. Hyperband: A Novel Bandit-Based Approach to Hyperparameter Optimization. J. Mach. Learn. Res. 18 (2017), 185:1–185:52. http://jmlr.org/papers/ v18/16-558.html

    work page 2017

  51. [51]

    Peng Li, Xi Rao, Jennifer Blase, Yue Zhang, Xu Chu, and Ce Zhang. 2021. CleanML: A Study for Evaluating the Impact of Data Cleaning on ML Classification Tasks. In 37th IEEE International Conference on Data Engineering, ICDE 2021, Chania, Greece, April 19-22, 2021. IEEE, 13–24. https://doi.org/10.1109/ICDE51399.2021.00009

    work page doi:10.1109/icde51399.2021.00009 2021

  52. [52]

    Yang Li, Yu Shen, Wentao Zhang, Jiawei Jiang, Yaliang Li, Bolin Ding, Jingren Zhou, Zhi Yang, Wentao Wu, Ce Zhang, and Bin Cui. 2021. VolcanoML: Speeding up End-to-End AutoML via Scalable Search Space Decomposition. Proc. VLDB Endow. 14, 11 (2021), 2167–2176. http://www.vldb.org/pvldb/vol14/p2167-li.pdf 15

    work page 2021

  53. [53]

    Marius Lindauer, Katharina Eggensperger, Matthias Feurer, André Biedenkapp, Difan Deng, Carolin Benjamins, René Sass, and Frank Hutter. 2021. SMAC3: A Versatile Bayesian Optimization Package for Hyperparameter Optimization. arXiv:2109.09831 [cs.LG]

    work page arXiv 2021

  54. [54]

    Chenxi Liu, Barret Zoph, Maxim Neumann, Jonathon Shlens, Wei Hua, Li-Jia Li, Li Fei-Fei, Alan Yuille, Jonathan Huang, and Kevin Murphy. 2018. Progressive neural architecture search. In Proceedings of the European conference on computer vision (ECCV). 19–34

    work page 2018

  55. [55]

    Sijia Liu, Parikshit Ram, Deepak Vijaykeerthy, Djallel Bouneffouf, Gregory Bram- ble, Horst Samulowitz, Dakuo Wang, Andrew Conn, and Alexander G. Gray

    work page

  56. [56]

    An ADMM Based Framework for AutoML Pipeline Configuration. In The Thirty-Fourth AAAI Conference on Artificial Intelligence, AAAI 2020, The Thirty- Second Innovative Applications of Artificial Intelligence Conference, IAAI 2020, The Tenth AAAI Symposium on Educational Advances in Artificial Intelligence, EAAI 2020, New York, NY, USA, February 7-12, 2020 . ...

    work page 2020

  57. [57]

    Lundberg and Su-In Lee

    Scott M. Lundberg and Su-In Lee. 2017. A Unified Approach to Interpreting Model Predictions. In Advances in Neural Information Processing Systems 30: Annual Conference on Neural Information Processing Systems 2017, December 4-9, 2017, Long Beach, CA, USA , Isabelle Guyon, Ulrike von Luxburg, Samy Bengio, Hanna M. Wallach, Rob Fergus, S. V. N. Vishwanathan...

    work page 2017

  58. [58]

    Shaul Markovitch and Dan Rosenstein. 2002. Feature Generation Using General Constructor Functions. Mach. Learn. 49, 1 (2002), 59–98. https://doi.org/10.1023/ A:1014046307775

    work page 2002

  59. [59]

    Microsoft. 2021. Neural Network Intelligence. https://github.com/microsoft/nni

    work page 2021

  60. [60]

    Kevin P Murphy. 2012. Machine learning: a probabilistic perspective . MIT press

    work page 2012

  61. [61]

    Khalil, and Deepak S

    Fatemeh Nargesian, Horst Samulowitz, Udayan Khurana, Elias B. Khalil, and Deepak S. Turaga. 2017. Learning Feature Engineering for Classification. In Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelli- gence, IJCAI 2017, Melbourne, Australia, August 19-25, 2017 , Carles Sierra (Ed.). ijcai.org, 2529–2535. https://doi.org/...

    work page doi:10.24963/ijcai.2017/352 2017

  62. [62]

    Randal S Olson and Jason H Moore. 2016. TPOT: A tree-based pipeline optimiza- tion tool for automating machine learning. In Workshop on automatic machine learning. PMLR, 66–74

    work page 2016

  63. [63]

    Efficient Neural Architecture Search via Parameter Sharing

    Hieu Pham, Melody Y. Guan, Barret Zoph, Quoc V. Le, and Jeff Dean. 2018. Efficient Neural Architecture Search via Parameter Sharing.CoRR abs/1802.03268 (2018). arXiv:1802.03268 http://arxiv.org/abs/1802.03268

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  64. [64]

    Alexandre Quemy. 2020. Two-stage optimization for machine learning workflow. Inf. Syst. 92 (2020), 101483. https://doi.org/10.1016/j.is.2019.101483

    work page doi:10.1016/j.is.2019.101483 2020

  65. [65]

    Esteban Real, Alok Aggarwal, Yanping Huang, and Quoc V. Le. 2018. Regularized Evolution for Image Classifier Architecture Search. CoRR abs/1802.01548 (2018). arXiv:1802.01548 http://arxiv.org/abs/1802.01548

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  66. [66]

    Nicolas Schilling, Martin Wistuba, Lucas Drumond, and Lars Schmidt-Thieme

    work page

  67. [67]

    Hyperparameter Optimization with Factorized Multilayer Perceptrons. In Machine Learning and Knowledge Discovery in Databases - European Confer- ence, ECML PKDD 2015, Porto, Portugal, September 7-11, 2015, Proceedings, Part II (Lecture Notes in Computer Science) , Annalisa Appice, Pedro Pereira Rodrigues, Vítor Santos Costa, João Gama, Alípio Jorge, and Ca...

    work page doi:10.1007/978-3-319-23525-7_6 2015

  68. [68]

    Nicolas Schilling, Martin Wistuba, and Lars Schmidt-Thieme. 2016. Scalable Hyperparameter Optimization with Products of Gaussian Process Experts. In Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2016, Riva del Garda, Italy, September 19-23, 2016, Proceedings, Part I (Lecture Notes in Computer Science) , Paolo Frasc...

    work page 2016

  69. [69]

    Vraj Shah, Jonathan Lacanlale, Premanand Kumar, Kevin Yang, and Arun Kumar

    work page

  70. [70]

    In SIGMOD ’21: International Conference on Management of Data, Virtual Event, China, June 20-25, 2021 , Guoliang Li, Zhanhuai Li, Stratos Idreos, and Divesh Srivastava (Eds.)

    Towards Benchmarking Feature Type Inference for AutoML Platforms. In SIGMOD ’21: International Conference on Management of Data, Virtual Event, China, June 20-25, 2021 , Guoliang Li, Zhanhuai Li, Stratos Idreos, and Divesh Srivastava (Eds.). ACM, 1584–1596. https://doi.org/10.1145/3448016.3457274

    work page doi:10.1145/3448016.3457274 2021

  71. [71]

    Zeyuan Shang, Emanuel Zgraggen, Benedetto Buratti, Ferdinand Kossmann, Philipp Eichmann, Yeounoh Chung, Carsten Binnig, Eli Upfal, and Tim Kraska

    work page

  72. [72]

    In Proceedings of the 2019 International Conference on Management of Data, SIG- MOD Conference 2019, Amsterdam, The Netherlands, June 30 - July 5, 2019, Peter A

    Democratizing Data Science through Interactive Curation of ML Pipelines. In Proceedings of the 2019 International Conference on Management of Data, SIG- MOD Conference 2019, Amsterdam, The Netherlands, June 30 - July 5, 2019, Peter A. Boncz, Stefan Manegold, Anastasia Ailamaki, Amol Deshpande, and Tim Kraska (Eds.). ACM, 1171–1188. https://doi.org/10.1145...

    work page doi:10.1145/3299869.3319863 2019

  73. [73]

    Qitao Shi, Ya-Lin Zhang, Longfei Li, Xinxing Yang, Meng Li, and Jun Zhou. 2020. SAFE: Scalable Automatic Feature Engineering Framework for Industrial Tasks. In 36th IEEE International Conference on Data Engineering, ICDE 2020, Dallas, TX, USA, April 20-24, 2020 . IEEE, 1645–1656. https://doi.org/10.1109/ICDE48307. 2020.00146

    work page doi:10.1109/icde48307 2020

  74. [74]

    Julien Siems, Lucas Zimmer, Arber Zela, Jovita Lukasik, Margret Keuper, and Frank Hutter. 2020. NAS-Bench-301 and the Case for Surrogate Benchmarks for Neural Architecture Search. CoRR abs/2008.09777 (2020). arXiv:2008.09777 https://arxiv.org/abs/2008.09777

    work page arXiv 2020

  75. [75]

    William R Thompson. 1933. On the likelihood that one unknown probability exceeds another in view of the evidence of two samples. Biometrika 25, 3-4 (1933), 285–294

    work page 1933

  76. [76]

    Chris Thornton, Frank Hutter, Holger H Hoos, and Kevin Leyton-Brown. 2013. Auto-WEKA: Combined selection and hyperparameter optimization of classifica- tion algorithms. In Proceedings of the 19th ACM SIGKDD international conference on Knowledge discovery and data mining . 847–855

    work page 2013

  77. [77]

    Chi Wang and Qingyun Wu. 2019. FLO: Fast and Lightweight Hyperparameter Optimization for AutoML. CoRR abs/1911.04706 (2019). arXiv:1911.04706 http: //arxiv.org/abs/1911.04706

    work page arXiv 2019

  78. [78]

    Ruoxi Wang, Bin Fu, Gang Fu, and Mingliang Wang. 2017. Deep & Cross Network for Ad Click Predictions. In Proceedings of the ADKDD’17, Halifax, NS, Canada, August 13 - 17, 2017 . ACM, 12:1–12:7. https://doi.org/10.1145/3124749.3124754

    work page doi:10.1145/3124749.3124754 2017

  79. [79]

    Tianxiang Wang, Jie Xu, and Jian-Qiang Hu. 2021. A Study on Efficient Com- puting Budget Allocation for a Two-Stage Problem. Asia Pac. J. Oper. Res. 38, 2 (2021), 2050044:1–2050044:20. https://doi.org/10.1142/S021759592050044X

    work page doi:10.1142/s021759592050044x 2021

  80. [80]

    Ronald J Williams. 1992. Simple statistical gradient-following algorithms for connectionist reinforcement learning. Machine learning 8, 3 (1992), 229–256

    work page 1992

Showing first 80 references.