The Neural Network for Online Learning Task Without Manual Feature Extraction

  • Yuriy Fedorenko
  • Valeriy Chernenkiy
  • Yuriy GapanyukEmail author
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 11554)


The article is devoted to the problem of feature extraction in online learning tasks. In many cases, the proper feature extraction is very time-consuming. Currently, in some cases, this problem is successfully solved by deep neural networks. However, deep models are computationally expensive and so hardly applicable for online learning tasks which require frequent updating of the model. This paper proposes the lightweight neural net architecture that can be learned in online mode and doesn’t require complex handcrafted features. The small sample processing time distinguishes the proposed model from more complex deep neural networks. The architecture and learning process of the proposed model are discussed in detail. The special attention is paid to fast software implementation. On benchmarks, we show that developed implementation processes one sample several times faster than implementations on the base of deep learning frameworks. The conducted experiments on CTR prediction task show that the proposed neural net with raw features gives the same performance as the logistic regression model with handcrafted features. For a clear description of the proposed architecture, we use the metagraph approach.


Feature extraction Online learning Sparse neural network CTR prediction Logistic loss ROC curve Metagraph 


  1. 1.
    Guyon, I., Gunn, S., Nikravesh, M., Zadeh, L.A. (eds.): Feature Extraction: Foundations and Applications. Studies in Fuzziness and Soft Computing. Springer, Heidelberg (2008)Google Scholar
  2. 2.
    Manning, C.D., Raghavan, P., Schutze, H.: Introduction to Information Retrieval. University Press, Cambridge (2010)Google Scholar
  3. 3.
    Peng, H., Long, F., Ding, C.: Feature selection based on mutual information: criteria of max-dependency, max-relevance, and min-redundancy. IEEE Trans. Pattern Anal. Mach. Intell. 27, 1226–1238 (2005)Google Scholar
  4. 4.
    LeCun, Y., Bengio, Y., Hinton, G.: Deep learning. Nature 521, 436–444 (2015)Google Scholar
  5. 5.
    Bishop, C.: Pattern Recognition and Machine Learning. Springer, Heidelberg (2006). Scholar
  6. 6.
    Ruder, S.: An Overview of Gradient Descent Optimization Algorithms (2016)Google Scholar
  7. 7.
  8. 8.
    Fedorenko, Y.S., Gapanyuk, Y.E.: The neural network with automatic feature selection for solving problems with categorical variables. In: Kryzhanovsky, B., Dunin-Barkowski, W., Redko, V., Tiumentsev, Y. (eds.) NEUROINFORMATICS 2018. SCI, vol. 799, pp. 129–135. Springer, Cham (2019). Scholar
  9. 9.
    Lasagne Documentation (2019).
  10. 10.
    PyTorch Documentation (2019).
  11. 11.
    Obuchowski, N.A., Bullen, J.A.: Receiver operating characteristic (ROC) curves: review of methods with applications in diagnostic medicine. Phys. Med. Biol. 63, 1361–1367 (2018)Google Scholar
  12. 12.
    He, X., et al.: Practical lessons from predicting clicks on ads at Facebook. In: Eighth International Workshop on Data Mining for Online Advertising, ADKDD 2014, pp. 1–9. ACM, New York (2014)Google Scholar
  13. 13.
    Fedorenko, Y.S., Gapanyuk, Y.E.: Multilevel neural net adaptive models using the metagraph approach. Opt. Mem. Neural Netw. 25(4), 228–235 (2016). Scholar
  14. 14.
    Fedorenko, Y.S., Gapanyuk, Y.E., Minakova, S.V.: The analysis of regularization in deep neural networks using metagraph approach. In: Kryzhanovsky, B., Dunin-Barkowski, W., Redko, V. (eds.) Advances in Neural Computation, Machine Learning, and Cognitive Research, pp. 3–8. Springer International Publishing, Cham (2018). Scholar

Copyright information

© Springer Nature Switzerland AG 2019

Authors and Affiliations

  • Yuriy Fedorenko
    • 1
  • Valeriy Chernenkiy
    • 1
  • Yuriy Gapanyuk
    • 1
    Email author
  1. 1.Bauman Moscow State Technical UniversityMoscowRussia

Personalised recommendations