KI - Künstliche Intelligenz

, Volume 29, Issue 4, pp 329–337 | Cite as

Beyond Manual Tuning of Hyperparameters

  • Frank Hutter
  • Jörg Lücke
  • Lars Schmidt-ThiemeEmail author
Technical Contribution


The success of hand-crafted machine learning systems in many applications raises the question of making machine learning algorithms more autonomous, i.e., to reduce the requirement of expert input to a minimum. We discuss two strategies towards this goal: (1) automated optimization of hyperparameters (including mechanisms for feature selection, preprocessing, model selection, etc) and (2) the development of algorithms with reduced sets of hyperparameters. Since many research directions (e.g., deep learning), show a tendency towards increasingly complex algorithms with more and more hyperparamters, the demand for both of these strategies continuously increases. We review recent hyperparameter optimization methods and discuss data-driven approaches to avoid the introduction of hyperparameters using unsupervised learning. We end in discussing how these complementary strategies can work hand-in-hand, representing a very promising approach towards autonomous machine learning.


Hyperparameter optimization Automatic machine learning Autonomous learning Deep learning 


  1. 1.
    Adams RP, Wallach HM, Ghahramani Z (2009) Learning the structure of deep sparse graphical models. ArXiv preprint, arXiv:1001.0160
  2. 2.
    Aha DW (1992) Generalizing from case studies: a case study. In: ML, pp 1–10Google Scholar
  3. 3.
    Bardenet R, Brendel M, Kégl B, Sebag M (2013) Collaborative hyperparameter tuning. In: Proceeidngs of ICML’13Google Scholar
  4. 4.
    Bengio Y (2000) Gradient-based optimization of hyperparameters. Neural Comput 12(8):1889–1900CrossRefGoogle Scholar
  5. 5.
    Bergstra J, Bardenet R, Bengio Y, Kégl B (2011) Algorithms for hyper-parameter optimization. In: Proceedings of NIPS’11Google Scholar
  6. 6.
    Bergstra J, Bengio Y (2012) Random search for hyper-parameter optimization. JMLR 13:281–305MathSciNetzbMATHGoogle Scholar
  7. 7.
    Bergstra J, Cox D (2013) Hyperparameter optimization and boosting for classifying facial expressions: How good can a “null” model be? ArXiv preprint, arXiv:1306.3476
  8. 8.
    Bergstra J, Yamins D, Cox DD (2013) Making a science of model search: Hyperparameter optimization in hundreds of dimensions for vision architectures. In: Proceedings of ICML’13Google Scholar
  9. 9.
    Berkes P, Turner R, Sahani M (2008) On sparsity and overcompleteness in image models. In: Proceedings of NIPS’08, vol 21Google Scholar
  10. 10.
    Blockeel H (2006) Experiment databases: a novel methodology for experimental research. In: Knowledge discovery in inductive databases, pp 72–85. SpringerGoogle Scholar
  11. 11.
    Brazdil P, Gama J, Henery B (1994) Characterizing the applicability of classification algorithms using meta-level learning. In: Proceedings of ECML’94, pp 83–102Google Scholar
  12. 12.
    Brochu E, Cora, V., de Freitas, N (2010) A tutorial on Bayesian optimization of expensive cost functions, with application to active user modeling and hierarchical reinforcement learning. ArXiv preprint, arXiv:1012.2599
  13. 13.
    Castiello C, Castellano G, Fanelli AM (2005) Meta-data: characterization of input features for meta-learning. In: Modeling decisions for artificial intelligence, pp 457–468. SpringerGoogle Scholar
  14. 14.
    Ciresan D, Meier U, Schmidhuber J (2012) Multi-column deep neural networks for image classification. In: Proceedings of CVPR’12, pp 3642–3649. IEEEGoogle Scholar
  15. 15.
    Dayan P (1997) Recognition in hierarchical models. In: Foundations of computational mathematics, pp 43–62. SpringerGoogle Scholar
  16. 16.
    Domhan T, Springenberg T, Hutter F (2014) Extrapolating learning curves of deep neural networks. In: ICML 2014 AutoML WorkshopGoogle Scholar
  17. 17.
    Eggensperger K, Feurer M, Hutter F, Bergstra J, Snoek J, Hoos H, Leyton-Brown K (2013) Towards an empirical foundation for assessing Bayesian optimization of hyperparameters. In: NIPS workshop on Bayesian Optimization in Theory and PracticeGoogle Scholar
  18. 18.
    Engels R, Theusinger C (1998) Using a data metric for preprocessing advice for data mining applications. In: Proceedings of ECAI’98, pp 430–434Google Scholar
  19. 19.
    Fawcett C, Hoos H (2013) Analysing differences between algorithm configurations through ablation. In: Proceedings of MIC’13, pp 123–132Google Scholar
  20. 20.
    Feurer M, Springenberg T, Hutter F (2015) Initializing Bayesian hyperparameter optimization via meta-learning. In: Proceedings of AAAI’15Google Scholar
  21. 21.
    Gomes TAF, Prudêncio RBC, Soares C, Rossi ALD (2012) Carvalho, A.C.P.L.F.: combining meta-learning and search techniques to select parameters for support vector machines. Neurocomputing 75(1):3–13CrossRefGoogle Scholar
  22. 22.
    Goodfellow I, Courville AC, Bengio Y (2012) Large-scale feature learning with spike-and-slab sparse coding. In: Proceedings of ICML’12Google Scholar
  23. 23.
    Griffiths TL, Kemp C, Tenenbaum JB (2008) Bayesian models of cognition. In: Sun R (ed) Cambridge Handbook of Computational Psychology. Cambridge University Press, New York, NY, USAGoogle Scholar
  24. 24.
    Gross S, Mokbel B, Hammer B, Pinkwart N (2012) Feedback provision strategies in intelligent tutoring systems based on clustered solution spaces. In: Desel J, Haake JM, Spannagel C (eds) DeLFI 2012: Die 10. e-Learning Fachtagung Informatik, pp 27–38. Köllen, Hagen, GermanyGoogle Scholar
  25. 25.
    Guerra SB, Prudłncio RB, Ludermir TB (2008) Predicting the performance of learning algorithms using support vector machines as meta-regressors. In: Proceedings of ICANN’08, pp 523–532Google Scholar
  26. 26.
    Guo X, Yang J, Wu C, Wang C, Liang Y (2008) A novel LS-SVMs hyper-parameter selection based on particle swarm optimization. Neurocomputing 71(16):3211–3215CrossRefGoogle Scholar
  27. 27.
    Henery RJ (1994) Methods for comparison. In: Michie D, Spiegelhalter DJ, Taylor CC (eds) Machine learning, neural and statistical classification. Ellis Horwood, New YorkGoogle Scholar
  28. 28.
    Hinton GE, Osindero S, Teh Y (2006) A fast learning algorithm for deep belief nets. Neural Comput 18(7)Google Scholar
  29. 29.
    Ho TK, Basu M (2002) Complexity measures of supervised classification problems. IEEE Trans Pattern Anal Mach Intell 24(3):289–300CrossRefGoogle Scholar
  30. 30.
    Hutter F, Hoos H, Leyton-Brown K (2014) An efficient approach for assessing hyperparameter importance. In: Proceeding of ICML’14, pp 754–762Google Scholar
  31. 31.
    Hutter F, Hoos H, Leyton-Brown K, Stützle T (2009) ParamILS: an automatic algorithm configuration framework. JAIR 36(1):267–306zbMATHGoogle Scholar
  32. 32.
    Hutter F, Hoos HH, Leyton-Brown K (2011) Sequential model-based optimization for general algorithm configuration. In: Proceeidngs of LION-5Google Scholar
  33. 33.
    Hutter F, Hoos HH, Leyton-Brown K (2013) Identifying key algorithm parameters and instance features using forward selection. In: Proceedings of LION-7Google Scholar
  34. 34.
    Jones DR, Schonlau M, Welch WJ (1998) Efficient global optimization of expensive black box functions. Journal of Global Optim 13:455–492MathSciNetCrossRefzbMATHGoogle Scholar
  35. 35.
    King RD, Feng C, Sutherland A (1995) Statlog: comparison of classification algorithms on large real-world problems. Appl Artif Intell 9(3):289–333CrossRefGoogle Scholar
  36. 36.
    Kingma DP, Mohamed S, Rezende DJ, Welling M (2014) Semi-supervised learning with deep generative models. In: Proceedings of NIPS’14, pp 3581–3589Google Scholar
  37. 37.
    Kulick J, Toussaint M, Lang T, Lopes M (2013) Active learning for teaching a robot grounded relational symbols. In: Proceedings of IJCAI’13Google Scholar
  38. 38.
    LeCun Y, Bengio Y (1995) Convolutional networks for images, speech, and time series. Handb Brain Theory Neural Netw 3361:310Google Scholar
  39. 39.
    LeCun Y, Bottou L, Bengio Y, Haffner P (2001) Gradient-based learning applied to document recognition. In: Intelligent Signal Processing, pp 306–351. IEEE PressGoogle Scholar
  40. 40.
    LeCun Y, Kavukcuoglu K, Farabet C (2010) Convolutional networks and applications in vision. Proceeidngs of ISCAS’10 pp 253–6 (2010)Google Scholar
  41. 41.
    Lee TS, Mumford D (2003) Hierarchical Bayesian inference in the visual cortex. J Opt Soc Am A Opt Image Sci Vis 20(7):1434–1448CrossRefGoogle Scholar
  42. 42.
    Lemke C, Budka M, Gabrys B (2013) Metalearning: a survey of trends and technologies. Artif. Intell. Rev. pp 1–14Google Scholar
  43. 43.
    Lücke J, Sahani M (2008) Maximal causes for non-linear component extraction. JMLR 9:1227–67zbMATHGoogle Scholar
  44. 44.
    Maron O, Moore A (1994) Hoeffding races: accelerating model selection search for classification and function approximation. In: Proceeding of NIPS’94, pp 59–66Google Scholar
  45. 45.
    Martius G, Der R, Ay N (2013) Information driven self-organization of complex robotic behaviors. PLoS One 8(5), e63,400. DOI 10.1371/journal.pone.0063400
  46. 46.
    Mohamed S, Heller K, Ghahramani Z (2012) Evaluating Bayesian and L1 approaches for sparse unsupervised learning. In: Proceedings of ICML’12Google Scholar
  47. 47.
    Murray I, Adams RP (2010) Slice sampling covariance hyperparameters of latent Gaussian models. In: Proceedings of NIPS’10, pp 1723–1731Google Scholar
  48. 48.
    Pasemann F (2013) Self-regulating neurons in the sensorimotor loop. In: Rojas I, Joya G, Gabestany J (eds) Advances in Computational Intelligence, vol 7902., Lecture Notes in Computer ScienceSpringer, Berlin Heidelberg, pp 481–491CrossRefGoogle Scholar
  49. 49.
    Peng Y, Flach PA, Brazdil P, Soares C (2002) Decision tree-based data characterization for meta-learning. In: ECML/PKDD’02 Workshop on Integration and Collaboration Aspects of Data Mining, Decision Support and Meta-Learning, pp 111–122Google Scholar
  50. 50.
    Pfahringer B, Bensusan H, Giraud-Carrier C (2000) Meta-learning by landmarking various learning algorithms. In: Proceedings of ICML’00, pp 743–750Google Scholar
  51. 51.
    Pinto F, Soares C, Mendes-Moreira J (2014) A framework to decompose and develop metafeatures. In: ECAI 2014 Workshop on Meta-Learning and Algorithm Selection, p 32Google Scholar
  52. 52.
    Reif M (2012) A comprehensive dataset for evaluating approaches of various meta-learning tasks. In: Proceedings of ICPRAM’12, vol 1, pp 273–276Google Scholar
  53. 53.
    Reif M, Shafait F, Dengel A (2012) Meta-learning for evolutionary parameter optimization of classifiers. Mach Learn 87(3):357–380MathSciNetCrossRefGoogle Scholar
  54. 54.
    Schmidhuber J (2015) Deep learning in neural networks: an overview. Neural Netw 61:85–117CrossRefGoogle Scholar
  55. 55.
    Schonlau M, Welch WJ, Jones DR (1998) Global versus local search in constrained optimization of computer models. In: New developments and applications in experimental design, vol 34, pp 11–25. Institute of Mathematical Statistics, Hayward, CaliforniaGoogle Scholar
  56. 56.
    Sheikh AS, Shelton JA, Lücke J (2014) A truncated em approach for spike-and-slab sparse coding. JMLR 15:2653–2687zbMATHGoogle Scholar
  57. 57.
    Sidenbladh H, Black MJ, Fleet DJ (2000) Stochastic tracking of 3d human figures using 2d image motion. In: Proceedings of the European Conference on Computer Vision (ECCV), pp 702–718. SpringerGoogle Scholar
  58. 58.
    Smith MR, Mitchell L, Giraud-Carrier C, Martinez T (2014) Recommending learning algorithms and their associated hyperparameters. ArXiv preprint, arXiv:1407.1890
  59. 59.
    Smith MR, White A, Giraud-Carrier C, Martinez T (2014) An easy to use repository for comparing and improving machine learning algorithm usage. ArXiv preprint, arXiv:1405.7292
  60. 60.
    Smith-Miles K (2009) Cross-disciplinary perspectives on meta-learning for algorithm selection. ACM Computing Surveys 41(1), 6:1–6:25Google Scholar
  61. 61.
    Snoek J, Larochelle H, Adams R (2012) Practical Bayesian optimization of machine learning algorithms. In: Proceedings of NIPS’12Google Scholar
  62. 62.
    Srinivas N, Krause A, Kakade S, Seeger M (2010) Gaussian process optimization in the bandit setting: No regret and experimental design. In: Proceedings of ICML’10Google Scholar
  63. 63.
    Srivastava N, Hinton G, Krizhevsky A, Sutskever I, Salakhutdinov R (2014) Dropout: a simple way to prevent neural networks from overfitting. JMLR 15(1):1929–1958MathSciNetzbMATHGoogle Scholar
  64. 64.
    Swersky K, Duvenaud D, Snoek J, Hutter F, Osborne M (2013) Raiders of the lost architecture: Kernels for Bayesian optimization in conditional parameter spaces. In: NIPS workshop on Bayesian Optimization in theory and practice (BayesOpt’13)Google Scholar
  65. 65.
    Swersky K, Snoek J, Adams R (2013) Multi-task bayesian optimization. In: Proc. of ICML’13Google Scholar
  66. 66.
    Swersky K, Snoek J, Prescott Adams R (2014) Freeze-Thaw Bayesian Optimization. ArXiv, arXiv:1406.3896
  67. 67.
    Thornton C, Hutter F, Hoos HH, Leyton-Brown K (2013) Auto-WEKA: combined selection and hyperparameter optimization of classification algorithms. In: Proceedings of KDD’13Google Scholar
  68. 68.
    Vanschoren J, Blockeel H, Pfahringer B, Holmes G (2012) Experiment databases: a new way to share, organize and learn from experiments. Machine Learning 87(2):127–158MathSciNetCrossRefzbMATHGoogle Scholar
  69. 69.
    Vilalta R, Drissi Y (2002) A perspective view and survey of meta-learning. Artif Intell Rev 18(2):77–95CrossRefGoogle Scholar
  70. 70.
    Vincent P, Larochelle H, Lajoie I, Bengio Y, Manzagol P (2010) Stacked denoising autoencoders: learning useful representations in a deep network with a local denoising criterion. JMLR 11:3371–3408MathSciNetzbMATHGoogle Scholar
  71. 71.
    Wager S, Wang S, Liang PS (2013) Dropout training as adaptive regularization. In: Proceedings of NIPS’13, pp 351–359Google Scholar
  72. 72.
    Weng P, Busa-Fekete R, Hüllermeier E (2013) Interactive Q-learning with ordinal rewards and unreliable tutor. In: Proceedings ECML/PKDD Workshop on Reinforcement learning from Generalized Feedback: Beyond Numerical RewardsGoogle Scholar
  73. 73.
    Yogatama D, Mann G (2014) Efficient transfer learning method for automatic hyperparameter tuning. In: Proceedings of AISTATS’14, pp 1077–1085Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2015

Authors and Affiliations

  • Frank Hutter
    • 1
  • Jörg Lücke
    • 2
  • Lars Schmidt-Thieme
    • 3
    Email author
  1. 1.University of FreiburgFreiburg im BreisgauGermany
  2. 2.Department for Medical Physics and Acoustics, Cluster of Excellence Hearing4allCarl von Ossietzky University OldenburgOldenburgGermany
  3. 3.University of HildesheimHildesheimGermany

Personalised recommendations