Post

AutoML Applications, Advantages and Top 30 Libraries ๐Ÿ“š

AutoML transforms ML workflows, boosting efficiency and innovation.


Application in Data Science

1. Exploratory Data Analysis (EDA)

  • โญ’ Initial Model Benchmarking
  • โญ’ Feature Importance

2. Model Development

  • โญ’ Rapid Prototyping
  • โญ’ Hyperparameter Tuning

3. Model Selection

  • โญ’ Comparison of Algorithms

4. Deployment and Production

  • โญ’ Automated Pipeline Creation
  • โญ’ Continuous Model Improvement

Key Advantages

  • โœบ Rapid Development: Accelerates the creation and deployment of models.

  • โœบ Accessibility: Makes machine learning approachable for non-experts.

  • โœบ Optimized Performance: Enhances model efficacy.

  • โœบ Error Reduction: Decreases manual errors in model selection and tuning.

  • โœบ Scalability & Resource Efficiency: Manages large datasets and optimizes computational resources.

  • โœบ Innovation: Explores more configurations and algorithms.

  • โœบ Adaptability: Quickly adjusts to new data, ensuring sustained model performance.


Python Libraries

๐Ÿ“šAuto-Sklearn

Automates scikit-learn model selection and training.

๐Ÿ“šTPOT

Uses genetic algorithms to optimize machine learning pipelines.

๐Ÿ“šH2O AutoML

Automates all steps of the machine learning process.

๐Ÿ“šAutoKeras

Simplifies the use of deep learning models.

๐Ÿ“šMLBox

Fast data reading and distributed preprocessing.

๐Ÿ“šAuto-ViML

Minimal coding for high-performance models.

๐Ÿ“šTransmogrifAI

Automates workflows on Apache Spark.

๐Ÿ“šFLAML

Finds accurate models with minimal computational cost.

๐Ÿ“šLudwig

Builds models using configuration files, no coding required.

๐Ÿ“šGAMA

Flexible AutoML pipeline support.

๐Ÿ“šHyperoptSklearn

Hyperparameter tuning for sklearn pipelines.

๐Ÿ“šAutoGluon

Simplifies deep learning tasks.

๐Ÿ“šNNI

Comprehensive feature engineering and model tuning tools.

๐Ÿ“šPyCaret

Low-code library that automates the ML workflow.

๐Ÿ“šMerlion

Focuses on time series machine learning.

๐Ÿ“šZenML

Builds production-ready MLOps pipelines.

๐Ÿ“šAuto-PyTorch

Automated architecture and hyperparameter optimization for PyTorch.

๐Ÿ“šStatsforecast

Fast forecasting for statistical models.

๐Ÿ“šAdanet

Automates learning of neural networks.

๐Ÿ“šIgel

Enables model operations without coding.

๐Ÿ“šMLJAR Supervised

Automates ML tasks on tabular data.

๐Ÿ“šKeras Tuner

Hyperparameter tuning for Keras.

๐Ÿ“šLazy Predict

Quickly builds and compares basic models.

๐Ÿ“šSparseML

Creates smaller, faster neural networks.

๐Ÿ“šPySR

Python and Julia-based symbolic regression.

๐Ÿ“šMetarank

Personalizes rankings with machine learning.

๐Ÿ“šAutoML-Zero

Evolves ML algorithms from scratch.

๐Ÿ“šAutoML for Images

Specializes in automating image data workflows.

๐Ÿ“šNNI

Integrates with ML frameworks for comprehensive automation.

๐Ÿ“šAutoGL

Tackles graph-based problems with automation.

Translate to Korean

AutoML์€ ML ์›Œํฌํ”Œ๋กœ๋ฅผ ํ˜์‹ ํ•˜์—ฌ ํšจ์œจ์„ฑ๊ณผ ํ˜์‹ ์„ ์ด‰์ง„ํ•ฉ๋‹ˆ๋‹ค.


๋ฐ์ดํ„ฐ ๊ณผํ•™์˜ ์‘์šฉ

1. ํƒ์ƒ‰์  ๋ฐ์ดํ„ฐ ๋ถ„์„(EDA)

  • โญ’ ์ดˆ๊ธฐ ๋ชจ๋ธ ๋ฒค์น˜๋งˆํ‚น
  • โญ’ ๊ธฐ๋Šฅ ์ค‘์š”๋„

2. ๋ชจ๋ธ ๊ฐœ๋ฐœ

  • โญ’ ์‹ ์†ํ•œ ํ”„๋กœํ†  ํƒ€์ดํ•‘
  • โญ’ ํ•˜์ดํผํŒŒ๋ผ๋ฏธํ„ฐ ํŠœ๋‹

3. ๋ชจ๋ธ ์„ ํƒ

  • โญ’ ์•Œ๊ณ ๋ฆฌ์ฆ˜ ๋น„๊ต

4. ๋ฐฐํฌ ๋ฐ ์ƒ์‚ฐ

  • โญ’ ์ž๋™ํ™”๋œ ํŒŒ์ดํ”„๋ผ์ธ ์ƒ์„ฑ
  • โญ’ ์ง€์†์ ์ธ ๋ชจ๋ธ ๊ฐœ์„ 

ํ•ต์‹ฌ ์žฅ์ 

  • โœบ ์‹ ์†ํ•œ ๊ฐœ๋ฐœ: ๋ชจ๋ธ์˜ ์ƒ์„ฑ ๋ฐ ๋ฐฐํฌ๋ฅผ ๊ฐ€์†ํ™”ํ•ฉ๋‹ˆ๋‹ค.

  • โœบ ์ ‘๊ทผ์„ฑ: ๋น„์ „๋ฌธ๊ฐ€๋„ ๊ธฐ๊ณ„ ํ•™์Šต์— ์‰ฝ๊ฒŒ ์ ‘๊ทผํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

  • โœบ ์ตœ์ ํ™”๋œ ์„ฑ๋Šฅ: ๋ชจ๋ธ ํšจ์œจ์„ฑ์„ ํ–ฅ์ƒ์‹œํ‚ต๋‹ˆ๋‹ค.

  • โœบ ์˜ค๋ฅ˜ ๊ฐ์†Œ: ๋ชจ๋ธ ์„ ํƒ ๋ฐ ํŠœ๋‹์—์„œ ์ˆ˜๋™ ์˜ค๋ฅ˜๋ฅผ ์ค„์ž…๋‹ˆ๋‹ค.

  • โœบ ํ™•์žฅ์„ฑ ๋ฐ ๋ฆฌ์†Œ์Šค ํšจ์œจ์„ฑ: ๋Œ€๊ทœ๋ชจ ๋ฐ์ดํ„ฐ ์„ธํŠธ๋ฅผ ๊ด€๋ฆฌํ•˜๊ณ  ์ปดํ“จํŒ… ๋ฆฌ์†Œ์Šค๋ฅผ ์ตœ์ ํ™”ํ•ฉ๋‹ˆ๋‹ค.

  • โœบ ํ˜์‹ : ๋” ๋งŽ์€ ๊ตฌ์„ฑ๊ณผ ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ํƒ์ƒ‰ํ•ฉ๋‹ˆ๋‹ค.

  • โœบ ์ ์‘์„ฑ: ์ƒˆ๋กœ์šด ๋ฐ์ดํ„ฐ์— ๋น ๋ฅด๊ฒŒ ์ ์‘ํ•˜์—ฌ ์ง€์†์ ์ธ ๋ชจ๋ธ ์„ฑ๋Šฅ์„ ๋ณด์žฅํ•ฉ๋‹ˆ๋‹ค.


Python ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ

๐Ÿ“šAuto-Sklearn

scikit-learn ๋ชจ๋ธ ์„ ํƒ ๋ฐ ํ•™์Šต์„ ์ž๋™ํ™”ํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ“šTPOT

์œ ์ „ ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ์‚ฌ์šฉํ•˜์—ฌ ๊ธฐ๊ณ„ ํ•™์Šต ํŒŒ์ดํ”„๋ผ์ธ์„ ์ตœ์ ํ™”ํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ“šH2O AutoML

๊ธฐ๊ณ„ ํ•™์Šต ํ”„๋กœ์„ธ์Šค์˜ ๋ชจ๋“  ๋‹จ๊ณ„๋ฅผ ์ž๋™ํ™”ํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ“šAutoKeras

๋”ฅ ๋Ÿฌ๋‹ ๋ชจ๋ธ์˜ ์‚ฌ์šฉ์„ ๋‹จ์ˆœํ™”ํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ“šMLBox

๋น ๋ฅธ ๋ฐ์ดํ„ฐ ์ฝ๊ธฐ ๋ฐ ๋ถ„์‚ฐ ์ „์ฒ˜๋ฆฌ.

๐Ÿ“šAuto-ViML

๊ณ ์„ฑ๋Šฅ ๋ชจ๋ธ์„ ์œ„ํ•œ ์ตœ์†Œํ•œ์˜ ์ฝ”๋”ฉ.

๐Ÿ“šTransmogrifAI

Apache Spark์—์„œ ์›Œํฌํ”Œ๋กœ๋ฅผ ์ž๋™ํ™”ํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ“šFLAML

์ตœ์†Œํ•œ์˜ ๊ณ„์‚ฐ ๋น„์šฉ์œผ๋กœ ์ •ํ™•ํ•œ ๋ชจ๋ธ์„ ์ฐพ์Šต๋‹ˆ๋‹ค.

๐Ÿ“šLudwig

์ฝ”๋”ฉ์ด ํ•„์š” ์—†๋Š” ๊ตฌ์„ฑ ํŒŒ์ผ์„ ์‚ฌ์šฉํ•˜์—ฌ ๋ชจ๋ธ์„ ๋นŒ๋“œํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ“šGAMA

์œ ์—ฐํ•œ AutoML ํŒŒ์ดํ”„๋ผ์ธ ์ง€์›.

๐Ÿ“šHyperoptSklearn

sklearn ํŒŒ์ดํ”„๋ผ์ธ์— ๋Œ€ํ•œ ํ•˜์ดํผ ๋งค๊ฐœ ๋ณ€์ˆ˜ ํŠœ๋‹.

๐Ÿ“šAutoGluon

๋”ฅ๋Ÿฌ๋‹ ์ž‘์—…์„ ๋‹จ์ˆœํ™”ํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ“šMNI

ํฌ๊ด„์ ์ธ ๊ธฐ๋Šฅ ์—”์ง€๋‹ˆ์–ด๋ง ๋ฐ ๋ชจ๋ธ ํŠœ๋‹ ๋„๊ตฌ.

๐Ÿ“šPyCaret

ML ์›Œํฌํ”Œ๋กœ๋ฅผ ์ž๋™ํ™”ํ•˜๋Š” ๋กœ์šฐ ์ฝ”๋“œ ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ์ž…๋‹ˆ๋‹ค.

๐Ÿ“šMerlion

์‹œ๊ณ„์—ด ๊ธฐ๊ณ„ ํ•™์Šต์— ์ค‘์ ์„ ๋‘ก๋‹ˆ๋‹ค.

๐Ÿ“šZenML

ํ”„๋กœ๋•์…˜์— ๋ฐ”๋กœ ์‚ฌ์šฉํ•  ์ˆ˜ ์žˆ๋Š” MLOps ํŒŒ์ดํ”„๋ผ์ธ์„ ๋นŒ๋“œํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ“šAuto-PyTorch

PyTorch๋ฅผ ์œ„ํ•œ ์ž๋™ํ™”๋œ ์•„ํ‚คํ…์ฒ˜ ๋ฐ ํ•˜์ดํผํŒŒ๋ผ๋ฏธํ„ฐ ์ตœ์ ํ™”.

๐Ÿ“šStatsforecase

ํ†ต๊ณ„ ๋ชจ๋ธ์— ๋Œ€ํ•œ ๋น ๋ฅธ ์˜ˆ์ธก.

๐Ÿ“šAdanet

์‹ ๊ฒฝ๋ง ํ•™์Šต์„ ์ž๋™ํ™”ํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ“šIgel

์ฝ”๋”ฉ ์—†์ด ๋ชจ๋ธ ์ž‘์—…์„ ์ˆ˜ํ–‰ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

๐Ÿ“šMLJAR Supervised

ํ…Œ์ด๋ธ” ํ˜•์‹ ๋ฐ์ดํ„ฐ์— ๋Œ€ํ•œ ML ์ž‘์—…์„ ์ž๋™ํ™”ํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ“šKeras Tuner

Keras์— ๋Œ€ํ•œ ํ•˜์ดํผ ๋งค๊ฐœ ๋ณ€์ˆ˜ ํŠœ๋‹.

๐Ÿ“šLazy Predict

๊ธฐ๋ณธ ๋ชจ๋ธ์„ ๋น ๋ฅด๊ฒŒ ๋นŒ๋“œํ•˜๊ณ  ๋น„๊ตํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

๐Ÿ“šSparseML

๋” ์ž‘๊ณ  ๋น ๋ฅธ ์‹ ๊ฒฝ๋ง์„ ๋งŒ๋“ญ๋‹ˆ๋‹ค.

๐Ÿ“šPySR

Python ๋ฐ Julia ๊ธฐ๋ฐ˜ ๊ธฐํ˜ธ ํšŒ๊ท€.

๐Ÿ“šMetarank

๊ธฐ๊ณ„ ํ•™์Šต์œผ๋กœ ์ˆœ์œ„๋ฅผ ๊ฐœ์ธํ™”ํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ“šAutoML-์ œ๋กœ

ML ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ์ฒ˜์Œ๋ถ€ํ„ฐ ๋ฐœ์ „์‹œํ‚ต๋‹ˆ๋‹ค.

๐Ÿ“š์ด๋ฏธ์ง€์šฉ AutoML

์ด๋ฏธ์ง€ ๋ฐ์ดํ„ฐ ์›Œํฌํ”Œ๋กœ์šฐ ์ž๋™ํ™”๋ฅผ ์ „๋ฌธ์œผ๋กœ ํ•ฉ๋‹ˆ๋‹ค.

๐Ÿ“šMNI

ํฌ๊ด„์ ์ธ ์ž๋™ํ™”๋ฅผ ์œ„ํ•ด ML ํ”„๋ ˆ์ž„์›Œํฌ์™€ ํ†ตํ•ฉ๋ฉ๋‹ˆ๋‹ค.

๐Ÿ“šAutoGL

์ž๋™ํ™”๋ฅผ ํ†ตํ•ด ๊ทธ๋ž˜ํ”„ ๊ธฐ๋ฐ˜ ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•ฉ๋‹ˆ๋‹ค

This post is licensed under CC BY 4.0 by the author.