You are here: Home › Research › Metalearning / Automatic Machine Learning (AutoML)

Metalearning / Automatic Machine Learning (AutoML)

.
Members

PhD's involved (LIAAD): Pavel Brazdil (FEP, coord.), João Gama (FEP), Alípio Jorge (FCUP), J.Mendes-Moreira (FEUP)
Collaboration: Carlos Soares (FEUP), Rui Leite (FEP, LIAAD),  Salisu Abdulrahman (U.Kano, Nigeria)

Main objectives and activities:

  • Exploit metaknowledge / metalearning to aid the user in selecting the appropriate ML (or other) algorithm for a given problem.
  • Aid the user in selecting the appropriate workflow / pipeline of operations for the given problem.
    Select from a range of data pre-processing and machine learning methods and their hyper-parameter configurations.
  • Study methods for composing algorithm / pipeline (workflow) porfolios involving both inclusion of new algorithm / pipelines  and reduction of existing portfolios. 
  • Planning experiments to enrich the current portfolios of algorithms / pipelines.
.

Books:

Articles with >100 citations (Google Scholar (GS)):

  • PB Brazdil, C Soares, JP da Costa, Ranking learning algorithms: Using IBL and meta-learning on accuracy and time results, Machine Learning 50 (3), 251-277, 2003, 470+ cit. (Sept.'20)
  • P Brazdil, J Gama, B Henery, Characterizing the applicability of classification algorithms using meta-level learning, Machine Learning: ECML-94, 83-102, 1994, 208+ cit. (Sept.'20)
  • J Gama, P Brazdil, Characterization of classification algorithms, Progress in Artificial Intelligence, 189-200, 1995, 180+ cit. (Sept.'20)
  • C Giraud-Carrier, R Vilalta, P Brazdil, Introduction to the special issue on meta-learning, Machine learning 54 (3), 187-193, 2004, 180+ cit. (Sept.'20)
  • C Soares, PB Brazdil, P Kuba: A meta-learning method to select the kernel width in support vector regression, Machine Learning 54 (3), 195-209, 2004, 180+ cit. (Sept.'20)
  • PB Brazdil, C Soares, A comparison of ranking methods for classification algorithm selection, Machine Learning: ECML 2000, 63-75, 2000, 180+ cit. (Jan.'19)

  • Y Peng, P Flach, C Soares, P Brazdil: Improved dataset characterisation for meta-learning, Discovery Science, Springer, Vol. 2534, 141-152, 2002, 140+ cit. (Sept.'20)

  • C Soares, PB Brazdil: Zoomed Ranking: Selection of Classification Algorithms Based on Relevant Performance Information, LNCS 1910, 126-135, 2000, 120+ cit. (Sept.'20)

Recent Events (Workshops):


Other Selected Articles:

2018

  • Salisu M Abdulrahman, P Brazdil, J van Rijn and J Vanschoren: Speeding up Algorithm Selection using Average Ranking and Active Testing by Introducing Runtime, in Special Issue on Metalearning and Algorithm Selection, Machine learning 107 (1), 2018 
  • P Brazdil, C Giraud-Carrier: Metalearning and Algorithm Selection: Progress, State of the Art and Introduction to the 2017 Special Issue, in Special Issue on Metalearning and Algorithm Selection, Machine learning 107 (1), 1-14, 2018

2017

  • MV Cachada, SM Abdulrahman, and P Brazdil, Combining Feature and Algorithm Hyperparameter Selection using some Metalearning Methods, Proc. of Workshop AutoML 2017 assoc. with ECML/PKDD 2017, Macedonia
  • SM Abdulrahman, MV Cachada, and P Brazdil, Impact of Feature Selection and Algorithm Parametrization on Average Ranking Method via Metalearning, in European Congress on Computational Methods in Applied Sciences and Engineering, Springer, 2017

2016

  • Catarina F Oliveira, C Soares, A Jorge, Can metalearning be applied to transfer on heterogeneous datasets?, In procedings of HAIS 2016 - 11th International Conference on Hybrid Artificial Intelligence Systems, Seville, Spain, Abril, 2016.
  • J Kanda, A de Carvalho, E Hruschka, C Soares, P Brazdil, Metalearning to select the best metaheuristic for the Traveling Salesman Problem: A comparison of metafeatures, Neurocomputing, 2016
  • S Abdulrahman, P Brazdil: Effect of Incomplete Meta-dataset on Average Ranking Method, presented at Auto@ML at ICML-2016. 

2015

  • S Abdulrahman, P Brazdil, J van Rijn and J Vanschoren:  Algorithm Selection via Metalearning and Sample-based Active Testing,  Proc. of MetaSel-2015 workshop of ECML/PKDD-2015
  • J. van Rijn, S Abdulrahman, P Brazdil and J. Vanschoren: Fast Algorithm Selection using Learning Curves, Proc. of IDA-2015, Springer, 2015
  • F Pinto, C Soares, P Brazdil, Combining regression models and metaheuristics to optimize space allocation in the retail industry, Intelligent Data Analysis 19, 149-162, 2015

  • F Pinto, C Soares, J Mendes-Moreira: Pruning bagging ensembles with metalearning, International Workshop on Multiple Classifier Systems, 64-75

2014

  • S Abdulrahman and P Brazdil:  Measures for Combining Accuracy and Time for Meta-learning, Proc. of MetaSel-2014 workshop of ECAI-2014.
  • Fábio Pinto, C Soares and J Mendes-Moreira: A Framework To Decompose And Develop Metafeatures, Proc. of MetaSel-2014 workshop of ECAI-2014

2012-10

  • R Leite, P Brazdil, Joaquin Vanschoren: Selecting Classification Algorithms with Active Testing. in Proc. of MLDM 2012: 117-131, Springer
  • R Leite, P Brazdil: Active Testing Strategy to Predict the Best Classification Algorithm via Sampling and Metalearning. ECAI 2010: 309-314
  • P Brazdil, R Vilalta, Ch Giraud-Carrier, C Soares: Metalearning. Encyclopedia of Machine Learning 2010: 662-666

2004-05

  • R Leite, P. Brazdil, Predicting Relative Performance of Classifiers from Samples, Proc. of the 22nd Int. Conf. on Machine Learning (ICML-2005), ACM, 2005

.

Ph.D. Theses Completed:

  • Catarina Félix de Oliveira, Metalearning for multiple-domain Transfer Learning, PRODEI, FEUP, 2019
  • Salisu M Abdulrahman: Development of Support System for Workflow Design for Data Mining Problems that Exploit Meta-learning, MAPi, 2017
  • Rui Leite: Control of Process of Knowledge Extraction from Data and Data Mining, FCUP, Porto, 2008
  • Carlos Soares: Learning Ranking s of Learning Algorithms, 2004


M.Sc Theses Completed:

  • Maria João Ferreira, Automated workflow design for classifying documents, MADSAD/FEP, 2017
  • Miguel Cachada, Automatic Design of Machine Learning Workflows, MADSAD/FEP, 2017

.

Metalearning and Algorithm Selection Google Group