{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,10]],"date-time":"2026-07-10T06:36:58Z","timestamp":1783665418907,"version":"3.55.0"},"reference-count":31,"publisher":"Springer Science and Business Media LLC","issue":"9","license":[{"start":{"date-parts":[[2025,6,20]],"date-time":"2025-06-20T00:00:00Z","timestamp":1750377600000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2025,6,20]],"date-time":"2025-06-20T00:00:00Z","timestamp":1750377600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"funder":[{"name":"Institute of Information Theory and Automation of the Czech Academy of Sciences"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["SIViP"],"published-print":{"date-parts":[[2025,9]]},"abstract":"<jats:title>Abstract<\/jats:title>\n          <jats:p>It is unclear whether generative approaches can achieve state-of-the-art performance with supervised classification in high-dimensional feature spaces and extremely small datasets. In this paper, we propose a drop-in variational autoencoder (VAE) for the task of supervised learning using an extremely small train set (i.e., <jats:inline-formula>\n              <jats:alternatives>\n                <jats:tex-math>$$n=1,.., 5$$<\/jats:tex-math>\n                <mml:math xmlns:mml=\"http:\/\/www.w3.org\/1998\/Math\/MathML\">\n                  <mml:mrow>\n                    <mml:mi>n<\/mml:mi>\n                    <mml:mo>=<\/mml:mo>\n                    <mml:mn>1<\/mml:mn>\n                    <mml:mo>,<\/mml:mo>\n                    <mml:mo>.<\/mml:mo>\n                    <mml:mo>.<\/mml:mo>\n                    <mml:mo>,<\/mml:mo>\n                    <mml:mn>5<\/mml:mn>\n                  <\/mml:mrow>\n                <\/mml:math>\n              <\/jats:alternatives>\n            <\/jats:inline-formula> images per class). Drop-in classifiers form a usual alternative when traditional approaches to Few-Shot Learning cannot be used. The classification will be defined as a posterior probability density function and approximated by the variational principle. We perform experiments on a large variety of deep feature representations extracted from different layers of popular convolutional neural network (CNN) architectures. We also benchmark with modern classifiers, including Neural Tangent Kernel (NTK), Support Vector Machine (SVM) with NTK kernel and Neural Network Gaussian Process (NNGP). Results obtained indicate that the drop-in VAE classifier outperforms all the compared classifiers in the extremely small data regime.<\/jats:p>","DOI":"10.1007\/s11760-025-04376-1","type":"journal-article","created":{"date-parts":[[2025,6,20]],"date-time":"2025-06-20T08:46:48Z","timestamp":1750409208000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":3,"title":["Small-data image classification via drop-in variational autoencoder"],"prefix":"10.1007","volume":"19","author":[{"given":"Babak","family":"Mahdian","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Radim","family":"Nedbal","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2025,6,20]]},"reference":[{"key":"4376_CR1","unstructured":"Finn, C., Abbeel, P. & Levine, S. Model-agnostic meta-learning for fast adaptation of deep networks ( 2017)"},{"key":"4376_CR2","doi-asserted-by":"crossref","unstructured":"Kornblith, S., Shlens, J. & Le, Q.\u00a0V. Do better imagenet models transfer better? ( 2019)","DOI":"10.1109\/CVPR.2019.00277"},{"key":"4376_CR3","unstructured":"Snell, J., Swersky, K. & Zemel, R.\u00a0S. Prototypical networks for few-shot learning.  CoRR arXiv:1703.05175 ( 2017)"},{"key":"4376_CR4","unstructured":"Nichol, A., Achiam, J. & Schulman, J. On first-order meta-learning algorithms.  CoRR  abs\/1803.02999 ( 2018)"},{"key":"4376_CR5","first-page":"3133","volume":"15","author":"MF Delgado","year":"2014","unstructured":"Delgado, M.F., Cernadas, E., Barro, S.: Do we need hundreds of classifiers to solve real world classification problems? J. M. L. R. 15, 3133\u20133181 (2014)","journal-title":"J. M. L. R."},{"key":"4376_CR6","unstructured":"Dua, D. & Graff, C. UCI machine learning repository ( 2017)"},{"key":"4376_CR7","unstructured":"Olson, M., Wyner, A. & Berk, R. NeurIPS (ed.)  Modern neural networks generalize on small data sets. (ed. NeurIPS) , 123\u2013456 ( 2018)"},{"key":"4376_CR8","doi-asserted-by":"crossref","unstructured":"Neal, R.\u00a0M.  Priors for Infinite Networks, 29\u201353 ( Springer New York, 1996)","DOI":"10.1007\/978-1-4612-0745-0_2"},{"key":"4376_CR9","unstructured":"Lee, J. et\u00a0al. ICLR (ed.)  Deep neural networks as Gaussian processes. (ed. ICLR)  ICLR ( OpenReview.net, 2018)"},{"key":"4376_CR10","unstructured":"Jacot, A. & Hongler, C. NeurIPS (ed.)  Neural tangent kernel: Convergence and generalization in neural networks. (ed. NeurIPS) , 8580\u20138589 ( 2018)"},{"key":"4376_CR11","unstructured":"Arora, S. et\u00a0al. ICLR (ed.)  Harnessing the power of infinitely wide deep nets on small-data tasks. (ed. ICLR)  ICLR 2020, Ethiopia, April 26-30, 2020 ( 2020)"},{"key":"4376_CR12","unstructured":"Kingma, D.\u00a0P. & Welling, M. ICLR (ed.)  Auto-encoding variational bayes. (ed. ICLR) ( 2014)"},{"key":"4376_CR13","volume-title":"& Kautz, J","author":"A Vahdat","year":"2021","unstructured":"Vahdat, A.: & Kautz, J. A deep hierarchical variational autoencoder, Nvae (2021)"},{"key":"4376_CR14","unstructured":"Havtorn, J.\u00a0D., Frellsen, J., Hauberg, S. & Maal\u00f8e, L. ICML (ed.)  Hierarchical vaes know what they don\u2019t know. (ed. ICML) , Vol. 139, 4117\u20134128 ( PMLR, 2021)"},{"key":"4376_CR15","unstructured":"Razavi, A., van\u00a0den Oord, A. & Vinyals, O. Generating diverse high-fidelity images with vq-vae-2 ( 2019)"},{"key":"4376_CR16","unstructured":"Hoyos, A. & Rivera, M. Attentive vq-vae ( 2024)"},{"key":"4376_CR17","unstructured":"Li, Z. & Liu, H. Beta-vae has 2 behaviors: Pca or ica? ( 2023)"},{"key":"4376_CR18","unstructured":"Burgess, C.\u00a0P. et\u00a0al.: Understanding disentangling in $$\\beta $$-vae ( 2018)"},{"key":"4376_CR19","unstructured":"Blei, D.\u00a0M., Kucukelbir, A. & McAuliffe, J.\u00a0D.: Variational inference: A review for statisticians.  CoRR arXiv:1601.00670 (2016)"},{"key":"4376_CR20","volume-title":"Pattern Recognition and Machine Learning","author":"CM Bishop","year":"2006","unstructured":"Bishop, C.M.: Pattern Recognition and Machine Learning. Information Science and Statistics) ( Springer-Verlag, New York Inc, Secaucus, NJ, USA (2006)"},{"key":"4376_CR21","unstructured":"Murphy, K.\u00a0P. Machine learning : a probabilistic perspective (MIT Press, 2012)"},{"key":"4376_CR22","doi-asserted-by":"crossref","unstructured":"Jordan, M.I., Ghahramani, Z., Jaakkola, T.S., Saul, L.K.: An introduction to variational methods for graphical models. Mach. Learn. 37, 183\u2013233 (1999)","DOI":"10.1023\/A:1007665907178"},{"key":"4376_CR23","unstructured":"Higgins, I. et\u00a0al. ICLR (ed.)  beta-vae: Learning basic visual concepts with a constrained variational framework. (ed. ICLR)  ICLR ( 2017)"},{"key":"4376_CR24","unstructured":"Lucas, J., Tucker, G. & Norouzi, M. CoRR (ed.)  Don\u2019t blame the ELBO! A linear VAE perspective on posterior collapse. (ed. CoRR) , 9403\u20139413 ( 2019)"},{"key":"4376_CR25","first-page":"993","volume":"3","author":"DM Blei","year":"2003","unstructured":"Blei, D.M., Ng, A.Y., Jordan, M.I.: Latent Dirichlet allocation. J. Mach. Learn. Res. 3, 993\u20131022 (2003)","journal-title":"J. Mach. Learn. Res."},{"key":"4376_CR26","volume-title":"& Wildberger, J","author":"M Fil","year":"2021","unstructured":"Fil, M., Mesinovic, M., Morris, M.: & Wildberger, J. Challenges and extensions, Beta-vae reproducibility (2021)"},{"key":"4376_CR27","unstructured":"Novak, R. et\u00a0al. Neural tangents: Easy infinite nns in python.  CoRR ( 2019)"},{"key":"4376_CR28","doi-asserted-by":"publisher","first-page":"1054","DOI":"10.1016\/j.patcog.2012.09.022","volume":"46","author":"N Maci\u00e0","year":"2013","unstructured":"Maci\u00e0, N., Bernad\u00f3-Mansilla, E., Orriols-Puig, A., Ho, T.: Learner excellence biased by data set selection. Pattern Recognition 46, 1054\u20131066 (2013)","journal-title":"Pattern Recognition"},{"key":"4376_CR29","unstructured":"Su, J. & Wu, G. f-vaes: Improve vaes with conditional flows ( 2018)"},{"key":"4376_CR30","unstructured":"Harvey, W., Naderiparizi, S. & Wood, F. Conditional image generation by conditioning variational auto-encoders ( 2022)"},{"key":"4376_CR31","unstructured":"Ramchandran, S., Tikhonov, G., L\u00f6nnroth, O., Tiikkainen, P. & L\u00e4hdesm\u00e4ki, H. Learning conditional variational autoencoders with missing covariates ( 2022)"}],"container-title":["Signal, Image and Video Processing"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s11760-025-04376-1.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/article\/10.1007\/s11760-025-04376-1\/fulltext.html","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s11760-025-04376-1.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,7,3]],"date-time":"2025-07-03T14:47:40Z","timestamp":1751554060000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1007\/s11760-025-04376-1"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,6,20]]},"references-count":31,"journal-issue":{"issue":"9","published-print":{"date-parts":[[2025,9]]}},"alternative-id":["4376"],"URL":"https:\/\/doi.org\/10.1007\/s11760-025-04376-1","relation":{},"ISSN":["1863-1703","1863-1711"],"issn-type":[{"value":"1863-1703","type":"print"},{"value":"1863-1711","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,6,20]]},"assertion":[{"value":"20 November 2024","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"8 May 2025","order":2,"name":"revised","label":"Revised","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"9 June 2025","order":3,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"20 June 2025","order":4,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}}],"article-number":"766"}}