Projective Latent Interventions for Understanding and Fine-Tuning Classifiers

Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 12446)


High-dimensional latent representations learned by neural network classifiers are notoriously hard to interpret. Especially in medical applications, model developers and domain experts desire a better understanding of how these latent representations relate to the resulting classification performance. We present Projective Latent Interventions (PLIs), a technique for retraining classifiers by back-propagating manual changes made to low-dimensional embeddings of the latent space. The back-propagation is based on parametric approximations of \(t\)-distributed stochastic neighbourhood embeddings. PLIs allow domain experts to control the latent decision space in an intuitive way in order to better match their expectations. For instance, the performance for specific pairs of classes can be enhanced by manually separating the class clusters in the embedding. We evaluate our technique on a real-world scenario in fetal ultrasound imaging.


Latent space Non-linear embedding Image classification 



This work was supported by the State of Upper Austria (Human-Interpretable Machine Learning) and the Austrian Federal Ministry of Education, Science and Research via the Linz Institute of Technology (LIT-2019-7-SEE-117), and by the Wellcome Trust (IEH 102431 and EPSRC EP/S013687/1.).


  1. 1.
    Baumgartner, C.F., et al.: SonoNet: real-time detection and localisation of fetal standard scan planes in freehand ultrasound. IEEE Trans. Med. Imaging 36(11), 2204–2215 (2017). Scholar
  2. 2.
    Bellet, A., Habrard, A., Sebban, M.: A survey on metric learning for feature vectors and structured data. arXiv preprint arXiv:1306.6709 (2013)
  3. 3.
    Brent, R.P.: Algorithms for minimization without derivatives. Courier Corporation (2013)Google Scholar
  4. 4.
    Chen, X., Weng, J., Lu, W., Xu, J., Weng, J.: Deep manifold learning combined with convolutional neural networks for action recognition. IEEE Trans. Neural Netw. Learn. Syst. 29(9), 3938–3952 (2017). Scholar
  5. 5.
    Dong, W., Moses, C., Li, K.: Efficient k-nearest neighbor graph construction for generic similarity measures. In: Proceedings of the 20th International Conference on World Wide Web, pp. 577–586 (2011).
  6. 6.
    Erhan, D., Bengio, Y., Courville, A., Manzagol, P.A., Vincent, P., Bengio, S.: Why does unsupervised pre-training help deep learning? J. Mach. Learn. Res. 11, 625–660 (2010).
  7. 7.
    Krizhevsky, A., Nair, V., Hinton, G.: CIFAR-10 (Canadian Institute for Advanced Research). Accessed 16 Mar 2020
  8. 8.
    Kulis, B., et al.: Metric learning: a survey. Found. Trends Mach. Learn. 5(4), 287–364 (2012)MathSciNetCrossRefGoogle Scholar
  9. 9.
    LeCun, Y., Cortes, C.: The MNIST database of handwritten digits (2005). Accessed 16 Mar 2020
  10. 10.
    Lee, C.Y., Xie, S., Gallagher, P., Zhang, Z., Tu, Z.: Deeply-supervised nets. In: Artificial intelligence and statistics, pp. 562–570 (2015).
  11. 11.
    van der Maaten, L.: Learning a parametric embedding by preserving local structure. In: Artificial Intelligence and Statistics, pp. 384–391 (2009).
  12. 12.
    van der Maaten, L., Hinton, G.: Visualizing data using t-SNE. J. Mach. Learn. Res. 9, 2579–2605 (2008).
  13. 13.
    McInnes, L., Healy, J., Melville, J.: UMAP: uniform manifold approximation and projection for dimension reduction, December 2018. arXiv:1802.03426
  14. 14.
    Mead, A.: Review of the development of multidimensional scaling methods. J. Roy. Stat. Soc. Ser. D (Stat.) 41(1), 27–39 (1992).
  15. 15.
    Min, M.R., van der Maaten, L., Yuan, Z., Bonner, A.J., Zhang, Z.: Deep supervised t-distributed embedding. In: Proceedings of the 27th International Conference on Machine Learning (ICML 2010) (2010).
  16. 16.
    NHS: Fetal Anomaly Screening Programme: Programme Handbook June 2015. Public Health England (2015)Google Scholar
  17. 17.
    Poličar, P.G., Stražar, M., Zupan, B.: openTSNE: a modular Python library for t-SNE dimensionality reduction and embedding. bioRxiv, August 2019.
  18. 18.
    Rauber, P.E., Fadel, S.G., Falcão, A.X., Telea, A.C.: Visualizing the hidden activity of artificial neural networks. IEEE Trans. Visual Comput. Graphics 23(1), 101–110 (2017)CrossRefGoogle Scholar
  19. 19.
    Rusu, A.A., et al.: Meta-learning with latent embedding optimization (2018). arXiv:1807.05960
  20. 20.
  21. 21.
    Tomar, V.S., Rose, R.C.: Manifold regularized deep neural networks. In: Fifteenth Annual Conference of the International Speech Communication Association (2014)Google Scholar
  22. 22.
    Wold, S., Esbensen, K., Geladi, P.: Principal component analysis. Chemometr. Intell. Lab. Syst. 2(1–3), 37–52 (1987). Scholar

Copyright information

© Springer Nature Switzerland AG 2020

Authors and Affiliations

  1. 1.Biomedical Image Analysis Group, Imperial CollegeLondonUK
  2. 2.Institute of Computer GraphicsJohannes Kepler University LinzLinzAustria

Personalised recommendations