University of Groningen
The snowball principle for handwritten word-image retrieval
van Oosten, Jean-Paul
DOI:
10.33612/diss.160750597
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2021
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
van Oosten, J-P. (2021). The snowball principle for handwritten word-image retrieval: The importance of labelled data and humans in the loop. University of Groningen. https://doi.org/10.33612/diss.160750597
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
“stellingen” — 2020/7/13 — 11:15 — page 1 — #1
Stellingen
behorende bij het proefschrift
The snowball principle for handwritten
word-image retrieval
van
Jean-Paul van Oosten
1. Trained hidden Markov models closer to the global optimum do not neces-sarily have a better maximum likelihood estimate (Chapter 2)
2. Without guidance, it is hard for hidden Markov models to converge to the global optimum (Chapter 2).
3. Hidden Markov models do not always find the actual stochastic transition structure in the data (Chapter 3).
4. Removing temporal information from hidden Markov models by flattening the transition matrix, surprisingly does not make them completely useless (Chapter 3).
5. Good recognizers are not by definition also good at ranking (Chapter 4). 6. Continuous labelling and updates to collections, requires us to regard
hand-writing recognition as a dynamic process (Chapter 4 and 5).
7. It is useful to study training algorithms that can end up in local optima from a global perspective.
8. A search engine and data mining platform for handwritten manuscripts is like a Tamagotchi: it needs constant attention and feeding.
9. Exploration is not just relevant to machine learning methods, it is also fun-damental to the methods of the research community as a whole: While Deep Learning has gained a lot of popularity, we should also venture outside of this local optimum.