Techniques for Automated Machine Learning

Qingquan Song; Xia Hu; Yi-Wei Chen

arxiv: 1907.08908 · v1 · pith:CO5RQKCGnew · submitted 2019-07-21 · 💻 cs.LG · cs.AI· stat.ML

Techniques for Automated Machine Learning

Yi-Wei Chen , Qingquan Song , Xia Hu This is my paper

Pith reviewed 2026-05-24 18:55 UTC · model grok-4.3

classification 💻 cs.LG cs.AIstat.ML

keywords automated machine learningAutoMLfeature engineeringhyperparameter optimizationdeep learningBayesian optimizationreinforcement learningevolutionary algorithms

0 comments

The pith

AutoML developments fall into automated feature engineering, model and hyperparameter learning, and deep learning, each drawing on Bayesian optimization, reinforcement learning, evolutionary algorithms, and gradient-based methods.

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

The paper organizes automated machine learning research into three categories: automated feature engineering, automated model and hyperparameter learning, and automated deep learning. It surveys the leading techniques within each category, including Bayesian optimization, reinforcement learning, evolutionary algorithms, and gradient-based approaches, along with existing frameworks. A reader would care because AutoML seeks to remove the need for extensive manual tuning so that domain experts can apply machine learning solutions directly.

Core claim

Automated machine learning aims to find optimal solutions automatically for a given problem. The field is reviewed by dividing current work into automated feature engineering, automated model and hyperparameter learning, and automated deep learning. State-of-the-art methods in these areas include Bayesian optimization, reinforcement learning, evolutionary algorithms, and gradient-based techniques. Popular frameworks are summarized and open challenges are noted.

What carries the argument

The three-way categorization of AutoML into automated feature engineering (AutoFE), automated model and hyperparameter learning (AutoMHL), and automated deep learning (AutoDL), which organizes the surveyed techniques.

If this is right

Domain experts without machine learning experience can access ready-to-use solutions.
Data scientists spend less time on manual tuning of features, models, and hyperparameters.
Open challenges identified at the end of the review point to remaining barriers in full automation.

Where Pith is reading between the lines

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

The three categories may need to be revisited as hybrid methods that span feature engineering and deep learning become more common.
Frameworks built on the surveyed techniques could be benchmarked against one another on shared tasks to reveal practical trade-offs.
Gradient-based approaches mentioned for deep learning may eventually influence the other two categories as well.

Load-bearing premise

The division of the field into exactly these three categories, together with the listed techniques and frameworks, adequately represents the state of AutoML when the paper was written.

What would settle it

A new technique or framework that cannot be placed in any of the three categories while still advancing the goal of automating machine learning pipelines would challenge the review's organizing structure.

read the original abstract

Automated machine learning (AutoML) aims to find optimal machine learning solutions automatically given a machine learning problem. It could release the burden of data scientists from the multifarious manual tuning process and enable the access of domain experts to the off-the-shelf machine learning solutions without extensive experience. In this paper, we review the current developments of AutoML in terms of three categories, automated feature engineering (AutoFE), automated model and hyperparameter learning (AutoMHL), and automated deep learning (AutoDL). State-of-the-art techniques adopted in the three categories are presented, including Bayesian optimization, reinforcement learning, evolutionary algorithm, and gradient-based approaches. We summarize popular AutoML frameworks and conclude with current open challenges of AutoML.

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.

Desk Editor's Note private letter to a colleague

2019 survey that maps AutoML into three categories and lists standard techniques but adds no new results or analysis.

read the letter

This is a 2019 survey on automated machine learning. It splits the area into automated feature engineering, automated model and hyperparameter learning, and automated deep learning, then walks through the usual methods—Bayesian optimization, reinforcement learning, evolutionary algorithms, gradient-based approaches—and names some frameworks at the time. The abstract and structure match what the reader described: a descriptive summary with no new algorithms or experiments. What it does reasonably well is lay out a simple organizational map and point to the main ideas that were current then, which can save a newcomer from starting from scratch. The soft spots are straightforward. It is purely a literature summary with no comparisons, no quantitative checks on the techniques, and no discussion of limitations or trade-offs. At this point the content is dated, so it does not reflect developments after 2019. The categorization is one reasonable way to group things but is not defended or shown to be superior to alternatives. This paper is for someone who wants a quick, high-level introduction to AutoML as it stood five years ago, such as a student or practitioner entering the area. It does not have the depth or currency for experts or for anyone needing current methods. I would bring it to a reading group only if the group is new to the topic and wants background. I would not cite it in my own work. It could go to peer review as a survey if a venue wants one, though it would need substantial updating to be useful now.

Referee Report

2 major / 2 minor

Summary. The paper reviews the current developments of AutoML, organizing the literature into three categories: automated feature engineering (AutoFE), automated model and hyperparameter learning (AutoMHL), and automated deep learning (AutoDL). It surveys state-of-the-art techniques including Bayesian optimization, reinforcement learning, evolutionary algorithms, and gradient-based approaches, summarizes popular frameworks, and identifies open challenges.

Significance. If the coverage is representative, the survey offers a structured synthesis of AutoML methods as of 2019 that could serve as an entry point for researchers. Its value is primarily organizational rather than generative, with no new derivations, empirical benchmarks, or machine-checked results to strengthen the contribution.

major comments (2)

[Introduction] The manuscript does not describe the literature review methodology (search strategy, databases, inclusion criteria, or cutoff date), which is load-bearing for the central claim that the selected techniques and frameworks represent current developments in the three categories.
[Conclusion] No quantitative validation or systematic comparison of the surveyed methods is provided, leaving the claim that the listed techniques are 'state-of-the-art' without supporting evidence of their relative performance or coverage.

minor comments (2)

The abstract and introduction could explicitly note the temporal scope of the review to help readers assess currency.
Some framework summaries would benefit from clearer pointers to the original papers or GitHub repositories for reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our survey of AutoML techniques. We address each major comment below.

read point-by-point responses

Referee: [Introduction] The manuscript does not describe the literature review methodology (search strategy, databases, inclusion criteria, or cutoff date), which is load-bearing for the central claim that the selected techniques and frameworks represent current developments in the three categories.

Authors: We agree that explicitly documenting the review process would improve transparency. In the revised manuscript we will add a paragraph to the Introduction describing the search strategy (keywords such as 'automated machine learning', 'AutoFE', 'Bayesian optimization for AutoML'), primary sources (Google Scholar, arXiv, and major conferences up to mid-2019), inclusion criteria (peer-reviewed or preprint works directly addressing the three categories), and the July 2019 cutoff date corresponding to the original submission. revision: yes
Referee: [Conclusion] No quantitative validation or systematic comparison of the surveyed methods is provided, leaving the claim that the listed techniques are 'state-of-the-art' without supporting evidence of their relative performance or coverage.

Authors: As a survey paper our contribution is organizational rather than empirical; we synthesize prominent methods from the 2019 literature without conducting new benchmarks. The 'state-of-the-art' designation reflects adoption and citation patterns in the cited works. We will partially revise the Conclusion to cite existing comparative studies and benchmark papers that evaluate subsets of these techniques, while noting that a full head-to-head evaluation lies outside the scope of this review. revision: partial

Circularity Check

0 steps flagged

No significant circularity: literature survey with no derivations

full rationale

This is a 2019 survey paper that partitions AutoML into AutoFE/AutoMHL/AutoDL categories and enumerates existing techniques (Bayesian optimization, RL, evolutionary algorithms, gradient-based methods) plus frameworks. The abstract and full text contain no equations, no predictions, no fitted parameters, and no deductive chain that could reduce to inputs by construction. The central content is descriptive enumeration of prior work; the categorization is presented as an organizational choice rather than a derived result. No self-citation load-bearing steps exist because no new theorem or prediction is advanced. This matches the default expectation of a non-circular review.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Review paper; contains no free parameters, axioms, or invented entities of its own.

pith-pipeline@v0.9.0 · 5643 in / 982 out tokens · 14791 ms · 2026-05-24T18:55:11.313551+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

36 extracted references · 36 canonical work pages · 1 internal anchor

[1]

Techniques for Automated Machine Learning

INTRODUCTION Automated machine learning (AutoML) has emerged as a prevailing research ﬁeld upon the ubiquitous adoption of machine learning techniques. It aims at automatically de- termining high-performance machine learning solutions with a little workforce in reasonable time budget. For example, Google HyperTune, Amazon Model Tuning, and Microsoft Azure...

work page internal anchor Pith review Pith/arXiv arXiv 2008
[2]

Order” entity based on the IDs of the“Customer

AUTOMA TED FEA TURE ENGINEERING Feature engineering is a process to manipulate features via operations such as data imputation, feature transforma- tion, and feature selection. It is a critical step in machine learning algorithms since suitable features directly inﬂuence their prediction performance [7]. Considering a datasetDF represented in terms of its...

work page
[3]

AUTOMA TED MODEL AND HYPERPA- RAMETER LEARNING Model and hyperparameter learning consists of model se- lectionandhyperparametertuning, whichoptimizesthepre- dictive performance by repeatedly changing machine learn- ingmodelsandtuningassociatehyperparametervalues. Given a dataset D divided into Dtrain and Dval for training and validationrespectively, wesho...

work page
[4]

It can be regarded as a particular topic of AutoMHL

AUTOMA TED DEEP LEARNING Automated deep learning (AutoDL) is purposed to facil- itate the design of neural architectures and the selection of their hyperparameters. It can be regarded as a particular topic of AutoMHL. Following Equation (6), we can describe AutoDL in the following optimization problem, A∗ = arg min A∈A Lval(A(w∗),Dval), s.t. w∗ = arg min ...

work page
[5]

Notethatweexcludethoseopen-sourcecodesthataremerely experiment implementations for research papers

AUTOML FRAMEWORKS In this section, we discuss representative AutoML frame- works from either open-source project or enterprise services. Notethatweexcludethoseopen-sourcecodesthataremerely experiment implementations for research papers. 5.1 Automated Feature Engineering FeatureTools[36]isanopen-sourceframeworkusingPython which can automatically generate f...

work page
[6]

Unfortunately, such benchmarks are not om- nipresent in AutoML

CHALLENGES Authoritative benchmarks are essential standard proto- colsforAutoMLcomparisonthatregulatedatasets, thesearch space, and the training setting of searched machine learn- ing solutions. Unfortunately, such benchmarks are not om- nipresent in AutoML. No standard protocol for AutoFE pro- vides a ﬁxed set of transformation operations, and identical ...

work page
[7]

Without laborious human involvement, AutoML can expedite the process of applying machine learning to speciﬁc domain problems

CONCLUSION Automated machine learning (AutoML) is the end-to-end process, which automatically discovers machine learning so- lutions. Without laborious human involvement, AutoML can expedite the process of applying machine learning to speciﬁc domain problems. Apart from our work, there are also other AutoML reviews [9, 4:], unveiling the mysteri- ous art ...

work page
[8]

J. S. Bergstra, R. Bardenet, Y. Bengio, and B. Kégl. Algorithms for hyper-parameter optimization. In Advances in neural information processing systems, pages 4768– 4776, 4233

work page
[9]

H. Cai, L. Zhu, and S. Han. ProxylessNAS: Direct neural architecture search on target task and hardware. InInternational Conference on Learning Representations, 423

work page
[10]

B. Chen, H. Wu, W. Mo, I. Chattopadhyay, and H. Lipson. Autostacker: A compositional evolutionary learning system. InProceedings of the Genetic and Evolutionary Computation Conference, GECCO ’3:, pages 624– 62;, New York, NY, USA,423:. ACM

work page
[11]

T. Chen, I. Goodfellow, and J. Shlens. Net4net: Accelerating learning via knowledge transfer.arXiv preprint arXiv:/one.oldstyle/five.oldstyle/one.oldstyle/one.oldstyle./zero.oldstyle/five.oldstyle/six.oldstyle/four.oldstyle/one.oldstyle, 4237

work page
[12]

P. M. Domingos. A few useful things to know about machine learning.Commun. acm, 77(32):9:– :9, 4234

work page
[13]

Eggensperger, M

K. Eggensperger, M. Feurer, F. Hutter, J. Bergstra, J. Snoek, H. Hoos, and K. Leyton-Brown. Towards an empirical foundation for assessing bayesian optimization of hyperparameters. InNIPS workshop on Bayesian Optimization in Theory and Practice, volume 32, page 5, 4235

work page
[14]

Elsken, J

T. Elsken, J. H. Metzen, and F. Hutter. Neural architecture search: A survey.Journal of Machine Learning Research, 42(77):3– 43, 423;. [:] S. Falkner, A. Klein, and F. Hutter. Bohb: Robust and eﬃcient hyperparameter optimization at scale. arXiv preprint arXiv:/one.oldstyle:/zero.oldstyle/seven.oldstyle./zero.oldstyle/one.oldstyle/seven.oldstyle/seven.olds...

work page
[15]

Hutter, H

F. Hutter, H. H. Hoos, and K. Leyton-Brown. Sequential model-based optimization for general algorithm conﬁguration. InInternational Conference on Learning and Intelligent Optimization, pages 729– 745. Springer, 4233

work page
[16]

Jamieson and A

K. Jamieson and A. Talwalkar. Non-stochastic best arm identiﬁcation and hyperparameter optimization. In Artiﬁcial Intelligence and Statistics, pages 462– 46:, 4238

work page
[17]

H. Jin, Q. Song, and X. Hu. Auto-keras: Eﬃcient neural architecture search with network morphism, 423:

work page
[18]

Kandasamy, W

K. Kandasamy, W. Neiswanger, J. Schneider, B. Poczos, and E. P. Xing. Neural architecture search with bayesian optimisation and optimal transport. In S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett, editors,Advances in Neural Information Processing Systems/three.oldstyle/one.oldstyle, pages 4238– 4247. Curran Associates, ...

work page
[19]

J. M. Kanter and K. Veeramachaneni. Deep feature synthesis: Towards automating data science endeavors. In /two.oldstyle/zero.oldstyle/one.oldstyle/five.oldstyleIEEE International Conference on Data Science and Advanced Analytics, DSAA/two.oldstyle/zero.oldstyle/one.oldstyle/five.oldstyle, Paris, France, October/one.oldstyle/nine.oldstyle-/two.oldstyle/one...

work page
[20]

G. Katz, E. C. R. Shin, and D. Song. Explorekit: Automatic feature generation and selection. In/two.oldstyle/zero.oldstyle/one.oldstyle/six.oldstyle IEEE /one.oldstyle/six.oldstyleth International Conference on Data Mining (ICDM), pages ;9;– ;:6, Dec 4238

work page
[21]

A. Kaul, S. Maheshwary, and V. Pudi. Autolearn: Automated feature generation and selection. In/two.oldstyle/zero.oldstyle/one.oldstyle/seven.oldstyle IEEE International Conference on Data Mining (ICDM), pages 439– 448, Nov 4239

work page
[22]

Khurana and H

U. Khurana and H. Samulowitz. Automating predictive modeling process using reinforcement learning. arXiv preprint arXiv:/one.oldstyle/nine.oldstyle/zero.oldstyle/three.oldstyle./zero.oldstyle/zero.oldstyle/seven.oldstyle/four.oldstyle/three.oldstyle, 423;. [3:] U. Khurana, H. Samulowitz, and D. Turaga. Feature engineering for predictive modeling using rei...

work page
[23]

Klein, S

A. Klein, S. Falkner, S. Bartels, P. Hennig, and F. Hutter. Fast bayesian optimization of machine learning hyperparameters on large datasets.arXiv preprint arXiv:/one.oldstyle/six.oldstyle/zero.oldstyle/five.oldstyle./zero.oldstyle/seven.oldstyle/zero.oldstyle/seven.oldstyle/nine.oldstyle, 4238

work page
[24]

L. Li, K. Jamieson, G. DeSalvo, A. Rostamizadeh, and A. Talwalkar. Hyperband: Bandit-based conﬁguration evaluation for hyperparameter optimization.4238

work page
[25]

H. Liu, K. Simonyan, and Y. Yang. DARTS: Diﬀerentiable architecture search. InInternational Conference on Learning Representations, 423

work page
[26]

R. Luo, F. Tian, T. Qin, E. Chen, and T.-Y. Liu. Neural architecture optimization. InAdvances in Neural Information Processing Systems, pages 9:48– 9:59, 423:

work page
[27]

Nickson, M

T. Nickson, M. A. Osborne, S. Reece, and S. J. Roberts. Automated machine learning on big data using stochastic algorithm tuning.arXiv preprint arXiv:/one.oldstyle/four.oldstyle/zero.oldstyle/seven.oldstyle./seven.oldstyle/nine.oldstyle/six.oldstyle/nine.oldstyle, 4236

work page
[28]

R. S. Olson, N. Bartley, R. J. Urbanowicz, and J. H. Moore. Evaluation of a tree-based pipeline optimization tool for automating data science. In Proceedings of the Genetic and Evolutionary Computation Conference /two.oldstyle/zero.oldstyle/one.oldstyle/six.oldstyle, GECCO ’38, pages 6:7– 6;4, New York, NY, USA,4238. ACM

work page
[29]

H. Pham, M. Guan, B. Zoph, Q. Le, and J. Dean. Eﬃcient neural architecture search via parameters sharing. In J. Dy and A. Krause, editors,Proceedings of the /three.oldstyle/five.oldstyleth International Conference on Machine Learning, volume :2 of Proceedings of Machine Learning Research, pages 62;7– 6326, StockholmsmÃďssan, Stockholm Sweden,32– 37 Jul 423:. PMLR

work page
[30]

F. Qi, Z. Xia, G. Tang, H. Yang, Y. Song, G. Qian, X. An, C. Lin, and G. Shi. Darwinml: A graph-based evolutionary algorithm for automated machine learning. arXiv preprint arXiv:/one.oldstyle/nine.oldstyle/zero.oldstyle/one.oldstyle./zero.oldstyle:/zero.oldstyle/one.oldstyle/three.oldstyle, 423:. [4:] Y. Quanming, W. Mengshuo, J. E. Hugo, G. Isabelle, H. ...

work page
[31]

B. Tran, B. Xue, and M. Zhang. Genetic programming for feature construction and selection in classiﬁcation on high-dimensional data.Memetic Computing, :(3):5– 37, 4238

work page
[32]

Weill, J

C. Weill, J. Gonzalvo, V. Kuznetsov, S. Yang, S. Yak, H. Mazzawi, E. Hotaj, G. Jerfel, V. Macko, B. Adlam, M. Mohri, and C. Cortes. Adanet: A scalable and ﬂexible framework for automatically learning ensembles, 423

work page
[33]

C. Ying, A. Klein, E. Real, E. Christiansen, K. Murphy, and F. Hutter. Nas-bench-323: Towards reproducible neural architecture search.arXiv preprint arXiv:/one.oldstyle/nine.oldstyle/zero.oldstyle/two.oldstyle./zero.oldstyle/nine.oldstyle/six.oldstyle/three.oldstyle/five.oldstyle, 423

work page
[34]

Yuanfei, W

L. Yuanfei, W. Mengshuo, Z. Hao, Y. Quanming, T. WeiWei, C. Yuqiang, Y. Qiang, and D. Wenyuan. Autocross: Automatic feature crossing for tabular data in real-world applications.arXiv preprint arXiv:/one.oldstyle/nine.oldstyle/zero.oldstyle/four.oldstyle./one.oldstyle/two.oldstyle:/five.oldstyle/seven.oldstyle, 423

work page
[35]

Zoph and Q

B. Zoph and Q. V. Le. Neural architecture search with reinforcement learning.arXiv preprint arXiv:/one.oldstyle/six.oldstyle/one.oldstyle/one.oldstyle./zero.oldstyle/one.oldstyle/five.oldstyle/seven.oldstyle:, 4238

work page
[36]

B. Zoph, V. Vasudevan, J. Shlens, and Q. V. Le. Learning transferable architectures for scalable image recognition. InThe IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 423:

work page

[1] [1]

Techniques for Automated Machine Learning

INTRODUCTION Automated machine learning (AutoML) has emerged as a prevailing research ﬁeld upon the ubiquitous adoption of machine learning techniques. It aims at automatically de- termining high-performance machine learning solutions with a little workforce in reasonable time budget. For example, Google HyperTune, Amazon Model Tuning, and Microsoft Azure...

work page internal anchor Pith review Pith/arXiv arXiv 2008

[2] [2]

Order” entity based on the IDs of the“Customer

AUTOMA TED FEA TURE ENGINEERING Feature engineering is a process to manipulate features via operations such as data imputation, feature transforma- tion, and feature selection. It is a critical step in machine learning algorithms since suitable features directly inﬂuence their prediction performance [7]. Considering a datasetDF represented in terms of its...

work page

[3] [3]

AUTOMA TED MODEL AND HYPERPA- RAMETER LEARNING Model and hyperparameter learning consists of model se- lectionandhyperparametertuning, whichoptimizesthepre- dictive performance by repeatedly changing machine learn- ingmodelsandtuningassociatehyperparametervalues. Given a dataset D divided into Dtrain and Dval for training and validationrespectively, wesho...

work page

[4] [4]

It can be regarded as a particular topic of AutoMHL

AUTOMA TED DEEP LEARNING Automated deep learning (AutoDL) is purposed to facil- itate the design of neural architectures and the selection of their hyperparameters. It can be regarded as a particular topic of AutoMHL. Following Equation (6), we can describe AutoDL in the following optimization problem, A∗ = arg min A∈A Lval(A(w∗),Dval), s.t. w∗ = arg min ...

work page

[5] [5]

Notethatweexcludethoseopen-sourcecodesthataremerely experiment implementations for research papers

AUTOML FRAMEWORKS In this section, we discuss representative AutoML frame- works from either open-source project or enterprise services. Notethatweexcludethoseopen-sourcecodesthataremerely experiment implementations for research papers. 5.1 Automated Feature Engineering FeatureTools[36]isanopen-sourceframeworkusingPython which can automatically generate f...

work page

[6] [6]

Unfortunately, such benchmarks are not om- nipresent in AutoML

CHALLENGES Authoritative benchmarks are essential standard proto- colsforAutoMLcomparisonthatregulatedatasets, thesearch space, and the training setting of searched machine learn- ing solutions. Unfortunately, such benchmarks are not om- nipresent in AutoML. No standard protocol for AutoFE pro- vides a ﬁxed set of transformation operations, and identical ...

work page

[7] [7]

Without laborious human involvement, AutoML can expedite the process of applying machine learning to speciﬁc domain problems

CONCLUSION Automated machine learning (AutoML) is the end-to-end process, which automatically discovers machine learning so- lutions. Without laborious human involvement, AutoML can expedite the process of applying machine learning to speciﬁc domain problems. Apart from our work, there are also other AutoML reviews [9, 4:], unveiling the mysteri- ous art ...

work page

[8] [8]

J. S. Bergstra, R. Bardenet, Y. Bengio, and B. Kégl. Algorithms for hyper-parameter optimization. In Advances in neural information processing systems, pages 4768– 4776, 4233

work page

[9] [9]

H. Cai, L. Zhu, and S. Han. ProxylessNAS: Direct neural architecture search on target task and hardware. InInternational Conference on Learning Representations, 423

work page

[10] [10]

B. Chen, H. Wu, W. Mo, I. Chattopadhyay, and H. Lipson. Autostacker: A compositional evolutionary learning system. InProceedings of the Genetic and Evolutionary Computation Conference, GECCO ’3:, pages 624– 62;, New York, NY, USA,423:. ACM

work page

[11] [11]

T. Chen, I. Goodfellow, and J. Shlens. Net4net: Accelerating learning via knowledge transfer.arXiv preprint arXiv:/one.oldstyle/five.oldstyle/one.oldstyle/one.oldstyle./zero.oldstyle/five.oldstyle/six.oldstyle/four.oldstyle/one.oldstyle, 4237

work page

[12] [12]

P. M. Domingos. A few useful things to know about machine learning.Commun. acm, 77(32):9:– :9, 4234

work page

[13] [13]

Eggensperger, M

K. Eggensperger, M. Feurer, F. Hutter, J. Bergstra, J. Snoek, H. Hoos, and K. Leyton-Brown. Towards an empirical foundation for assessing bayesian optimization of hyperparameters. InNIPS workshop on Bayesian Optimization in Theory and Practice, volume 32, page 5, 4235

work page

[14] [14]

Elsken, J

T. Elsken, J. H. Metzen, and F. Hutter. Neural architecture search: A survey.Journal of Machine Learning Research, 42(77):3– 43, 423;. [:] S. Falkner, A. Klein, and F. Hutter. Bohb: Robust and eﬃcient hyperparameter optimization at scale. arXiv preprint arXiv:/one.oldstyle:/zero.oldstyle/seven.oldstyle./zero.oldstyle/one.oldstyle/seven.oldstyle/seven.olds...

work page

[15] [15]

Hutter, H

F. Hutter, H. H. Hoos, and K. Leyton-Brown. Sequential model-based optimization for general algorithm conﬁguration. InInternational Conference on Learning and Intelligent Optimization, pages 729– 745. Springer, 4233

work page

[16] [16]

Jamieson and A

K. Jamieson and A. Talwalkar. Non-stochastic best arm identiﬁcation and hyperparameter optimization. In Artiﬁcial Intelligence and Statistics, pages 462– 46:, 4238

work page

[17] [17]

H. Jin, Q. Song, and X. Hu. Auto-keras: Eﬃcient neural architecture search with network morphism, 423:

work page

[18] [18]

Kandasamy, W

K. Kandasamy, W. Neiswanger, J. Schneider, B. Poczos, and E. P. Xing. Neural architecture search with bayesian optimisation and optimal transport. In S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett, editors,Advances in Neural Information Processing Systems/three.oldstyle/one.oldstyle, pages 4238– 4247. Curran Associates, ...

work page

[19] [19]

J. M. Kanter and K. Veeramachaneni. Deep feature synthesis: Towards automating data science endeavors. In /two.oldstyle/zero.oldstyle/one.oldstyle/five.oldstyleIEEE International Conference on Data Science and Advanced Analytics, DSAA/two.oldstyle/zero.oldstyle/one.oldstyle/five.oldstyle, Paris, France, October/one.oldstyle/nine.oldstyle-/two.oldstyle/one...

work page

[20] [20]

G. Katz, E. C. R. Shin, and D. Song. Explorekit: Automatic feature generation and selection. In/two.oldstyle/zero.oldstyle/one.oldstyle/six.oldstyle IEEE /one.oldstyle/six.oldstyleth International Conference on Data Mining (ICDM), pages ;9;– ;:6, Dec 4238

work page

[21] [21]

A. Kaul, S. Maheshwary, and V. Pudi. Autolearn: Automated feature generation and selection. In/two.oldstyle/zero.oldstyle/one.oldstyle/seven.oldstyle IEEE International Conference on Data Mining (ICDM), pages 439– 448, Nov 4239

work page

[22] [22]

Khurana and H

U. Khurana and H. Samulowitz. Automating predictive modeling process using reinforcement learning. arXiv preprint arXiv:/one.oldstyle/nine.oldstyle/zero.oldstyle/three.oldstyle./zero.oldstyle/zero.oldstyle/seven.oldstyle/four.oldstyle/three.oldstyle, 423;. [3:] U. Khurana, H. Samulowitz, and D. Turaga. Feature engineering for predictive modeling using rei...

work page

[23] [23]

Klein, S

A. Klein, S. Falkner, S. Bartels, P. Hennig, and F. Hutter. Fast bayesian optimization of machine learning hyperparameters on large datasets.arXiv preprint arXiv:/one.oldstyle/six.oldstyle/zero.oldstyle/five.oldstyle./zero.oldstyle/seven.oldstyle/zero.oldstyle/seven.oldstyle/nine.oldstyle, 4238

work page

[24] [24]

L. Li, K. Jamieson, G. DeSalvo, A. Rostamizadeh, and A. Talwalkar. Hyperband: Bandit-based conﬁguration evaluation for hyperparameter optimization.4238

work page

[25] [25]

H. Liu, K. Simonyan, and Y. Yang. DARTS: Diﬀerentiable architecture search. InInternational Conference on Learning Representations, 423

work page

[26] [26]

R. Luo, F. Tian, T. Qin, E. Chen, and T.-Y. Liu. Neural architecture optimization. InAdvances in Neural Information Processing Systems, pages 9:48– 9:59, 423:

work page

[27] [27]

Nickson, M

T. Nickson, M. A. Osborne, S. Reece, and S. J. Roberts. Automated machine learning on big data using stochastic algorithm tuning.arXiv preprint arXiv:/one.oldstyle/four.oldstyle/zero.oldstyle/seven.oldstyle./seven.oldstyle/nine.oldstyle/six.oldstyle/nine.oldstyle, 4236

work page

[28] [28]

R. S. Olson, N. Bartley, R. J. Urbanowicz, and J. H. Moore. Evaluation of a tree-based pipeline optimization tool for automating data science. In Proceedings of the Genetic and Evolutionary Computation Conference /two.oldstyle/zero.oldstyle/one.oldstyle/six.oldstyle, GECCO ’38, pages 6:7– 6;4, New York, NY, USA,4238. ACM

work page

[29] [29]

H. Pham, M. Guan, B. Zoph, Q. Le, and J. Dean. Eﬃcient neural architecture search via parameters sharing. In J. Dy and A. Krause, editors,Proceedings of the /three.oldstyle/five.oldstyleth International Conference on Machine Learning, volume :2 of Proceedings of Machine Learning Research, pages 62;7– 6326, StockholmsmÃďssan, Stockholm Sweden,32– 37 Jul 423:. PMLR

work page

[30] [30]

F. Qi, Z. Xia, G. Tang, H. Yang, Y. Song, G. Qian, X. An, C. Lin, and G. Shi. Darwinml: A graph-based evolutionary algorithm for automated machine learning. arXiv preprint arXiv:/one.oldstyle/nine.oldstyle/zero.oldstyle/one.oldstyle./zero.oldstyle:/zero.oldstyle/one.oldstyle/three.oldstyle, 423:. [4:] Y. Quanming, W. Mengshuo, J. E. Hugo, G. Isabelle, H. ...

work page

[31] [31]

B. Tran, B. Xue, and M. Zhang. Genetic programming for feature construction and selection in classiﬁcation on high-dimensional data.Memetic Computing, :(3):5– 37, 4238

work page

[32] [32]

Weill, J

C. Weill, J. Gonzalvo, V. Kuznetsov, S. Yang, S. Yak, H. Mazzawi, E. Hotaj, G. Jerfel, V. Macko, B. Adlam, M. Mohri, and C. Cortes. Adanet: A scalable and ﬂexible framework for automatically learning ensembles, 423

work page

[33] [33]

C. Ying, A. Klein, E. Real, E. Christiansen, K. Murphy, and F. Hutter. Nas-bench-323: Towards reproducible neural architecture search.arXiv preprint arXiv:/one.oldstyle/nine.oldstyle/zero.oldstyle/two.oldstyle./zero.oldstyle/nine.oldstyle/six.oldstyle/three.oldstyle/five.oldstyle, 423

work page

[34] [34]

Yuanfei, W

L. Yuanfei, W. Mengshuo, Z. Hao, Y. Quanming, T. WeiWei, C. Yuqiang, Y. Qiang, and D. Wenyuan. Autocross: Automatic feature crossing for tabular data in real-world applications.arXiv preprint arXiv:/one.oldstyle/nine.oldstyle/zero.oldstyle/four.oldstyle./one.oldstyle/two.oldstyle:/five.oldstyle/seven.oldstyle, 423

work page

[35] [35]

Zoph and Q

B. Zoph and Q. V. Le. Neural architecture search with reinforcement learning.arXiv preprint arXiv:/one.oldstyle/six.oldstyle/one.oldstyle/one.oldstyle./zero.oldstyle/one.oldstyle/five.oldstyle/seven.oldstyle:, 4238

work page

[36] [36]

B. Zoph, V. Vasudevan, J. Shlens, and Q. V. Le. Learning transferable architectures for scalable image recognition. InThe IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 423:

work page