frameworks.yaml

---
#for doc purpose using <placeholder:default_value> syntax when it applies.

# FORMAT:
__dummy_framework_with_defaults:
  version: ''
  module: # defaults to `frameworks.framework_name`
  setup_args: ''
  params: {}
  project: http://url/to/project/repo
  image: # will result in built image `author/image:tag`
    author: automlbenchmark
    image:  # defaults to `framework name to lowercase`
    tag:  # defaults to `framework version`


#########################
### AutoML frameworks ###
#########################

AutoGluon:
  version: "stable"
  description: |
    AutoGluon-Tabular: Unlike existing AutoML frameworks that primarily focus on model/hyperparameter selection,
    AutoGluon-Tabular succeeds by ensembling multiple models and stacking them in multiple layers.
  project: https://auto.gluon.ai
  refs: [https://arxiv.org/abs/2003.06505]
#  params:
#    _save_artifacts: ['leaderboard', 'models', 'info']

AutoGluon_bestquality:
  extends: AutoGluon
  description: |
    AutoGluon with 'best_quality' preset provides the most accurate overall predictor.
  params:
    presets: best_quality

autosklearn:
  version: 'stable'
  description: |
    auto-sklearn frees a machine learning user from algorithm selection and hyperparameter tuning.
    It leverages recent advantages in Bayesian optimization, meta-learning and ensemble construction.
  project: https://automl.github.io/auto-sklearn/
  refs: [http://papers.nips.cc/paper/5872-efficient-and-robust-automated-machine-learning.pdf]
#  params:
#    _save_artifacts: ['models', 'debug_as_files']
#    _n_jobs: 1

autosklearn2:
  extends: autosklearn
  params:
    _askl2: true

AutoWEKA:
  version: 'stable'
  description: |
    Auto-WEKA considers the problem of simultaneously selecting a learning algorithm and setting its hyperparameters, going beyond previous methods that address these issues in isolation.
    Auto-WEKA does this using a fully automated approach, leveraging recent innovations in Bayesian optimization.
  project: https://www.cs.ubc.ca/labs/beta/Projects/autoweka/
  refs: [https://www.cs.ubc.ca/labs/beta/Projects/autoweka/papers/16-599.pdf]

autoxgboost:
  version: 'latest'
  description: |
    autoxgboost aims to find an optimal xgboost model automatically using the machine learning framework mlr and the bayesian optimization framework mlrMBO.
  project: https://github.com/ja-thomas/autoxgboost
  refs: [https://arxiv.org/abs/1807.03873v2]

flaml:
  version: 'stable'
  description: |
    FLAML is a lightweight Python library that finds accurate machine learning models 
    automatically, efficiently and economically. It frees users from selecting learners 
    and hyperparameters for each learner. It is fast and cheap. 
  project: https://github.com/microsoft/FLAML
  refs: [https://arxiv.org/pdf/1911.04706.pdf]

GAMA:
  version: 'stable'
  description: |
    GAMA tries to find a good machine learning pipeline.
    For the machine learning pipeline GAMA considers data preprocessing steps, various machine learning algorithms, and their possible hyperparameters configurations.
  project: https://github.com/PGijsbers/gama
  refs: [https://joss.theoj.org/papers/10.21105/joss.01132]

H2OAutoML:
  version: 'stable'
  description: |
    H2O AutoML is a highly scalable, fully-automated, supervised learning algorithm
    which automates the process of training a large selection of candidate models and stacked ensembles within a singlefunction.
  project: http://docs.h2o.ai/h2o/latest-stable/h2o-docs/automl.html
  refs: [https://www.automl.org/wp-content/uploads/2020/07/AutoML_2020_paper_61.pdf]

hyperoptsklearn:
  version: 'latest'
  description: |
    hyperopt-sklearn uses Hyperopt to define a search space that encompasses many standard Scikit-Learn components and common patterns of composing them together.
  project: http://hyperopt.github.io/hyperopt-sklearn/
  refs: [http://conference.scipy.org/proceedings/scipy2014/pdfs/komer.pdf]
#  params:
#    max_evals: 1000
#    algo: hyperopt.tpe.suggest
#    verbose: true

lightautoml:
  project: https://github.com/sberbank-ai-lab/LightAutoML
  version: 'stable'
  description: |
    LightAutoML (LAMA) project from Sberbank AI Lab AutoML group is the framework for automatic classification and regression model creation.

mljarsupervised:
  version: 'stable'
  description: |
    AutoML mljar-supervised abstracts the common way to preprocess the data, construct the machine learning models, and perform hyper-parameters tuning to find the best model.
  project: https://supervised.mljar.com/
#  params:
#    algorithms: ["Baseline"]
#    _save_artifacts: True

mljarsupervised_compete:
  extends: mljarsupervised
  description: "MLJAR is using 'Compete' mode to provide the most accurate predictor"
  params:
    mode: Compete   # set mode for Compete, default mode is Explain
    
MLNet:
  version: 'latest'
  description: |
    MLNET.CLI is a automated machine learning tool implemented by ml.net.
    
MLPlan:
  version: 'stable'
  abstract: true
  description: |
    ML-Plan is an approach to AutoML based on hierarchical task networks (HTNs).
  project: http://mlplan.org
  refs: [https://doi.org/10.1007/s10994-018-5735-z]

MLPlanSKLearn:
  extends: MLPlan
  params:
    _backend: sklearn

MLPlanWEKA:
  extends: MLPlan
  params:
    _backend: weka

mlr3automl:
  version: 'stable'
  project: https://github.com/a-hanf/mlr3automl

oboe:
  version: 'latest'
  description: |
    oboe is an AutoML model selection system to select estimators or pipelines for a dataset.
  project: https://github.com/udellgroup/oboe
#  params:
#    build_ensemble: false
#    selection_method: random
#    verbose: true

TPOT:
  version: 'stable'
  description: |
    TPOT is a Python Automated Machine Learning tool that optimizes machine learning pipelines using genetic programming.
  project: https://github.com/EpistasisLab/tpot
  refs:
    - https://academic.oup.com/bioinformatics/article/36/1/250/5511404
    - http://doi.acm.org/10.1145/2908812.2908918
#  params:
#    _save_artifacts: ['models']
#    max_eval_time_mins: 2
#    population_size: 25
#    verbosity: 2



#######################################
### Non AutoML reference frameworks ###
#######################################

constantpredictor:
  description: |
    Fast dummy classifier mainly used to test the app and/or datasets: this doesn't encode data.
  project: https://scikit-learn.org/stable/modules/generated/sklearn.dummy.DummyClassifier.html

DecisionTree:
  version: 'stable'
  description: |
    A simple decision tree implementation (scikit-learn) used for testing or as a reference base.
  project: https://scikit-learn.org/stable/modules/generated/sklearn.tree.DecisionTreeClassifier.html

RandomForest:
  version: 'stable'
  description: |
    A simple Random Forest implementation (scikit-learn) used for testing or as a reference base.
  project: http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html
  params:
    n_estimators: 2000
#    _n_jobs: 1   # faster, fitter, happier (running OoM on some datasets when using multiprocessing)
#    verbose: true

ranger:
  version: 'stable'
  description: |
    A Random Forest implementation on R, mainly used for testing and as an example of R integration.
  project: https://github.com/imbs-hl/ranger

TunedRandomForest:
  version: 'stable'
  description: |
    A Tuned Random Forest implementation (scikit-learn) trying to optimize over max_features, and used as a reference base.
  project: http://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestClassifier.html
  params:
    n_estimators: 2000
#    _n_jobs: 1  # cf. RandomForest
#    _tuning:
#      n_estimators: 500