Semi-supervised deep learning in sequence labeling
Název práce v češtině: | Semisupervizované hluboké učení v označování sekvencí |
---|---|
Název v anglickém jazyce: | Semi-supervised deep learning in sequence labeling |
Klíčová slova: | hluboké učení, semi-supervizované učení, modelovaní sekvencí, označování sekvencí, strojové učení, neuronové sítě |
Klíčová slova anglicky: | deep learning, semi-supervised learning, sequence modeling, sequence labeling, machine learning, neural networks |
Akademický rok vypsání: | 2018/2019 |
Typ práce: | diplomová práce |
Jazyk práce: | angličtina |
Ústav: | Katedra teoretické informatiky a matematické logiky (32-KTIML) |
Vedoucí / školitel: | Tomáš Šabata |
Řešitel: | Mgr. Juraj Eduard Páll - zadáno a potvrzeno stud. odd. |
Datum přihlášení: | 05.03.2019 |
Datum zadání: | 11.03.2019 |
Datum potvrzení stud. oddělením: | 22.03.2019 |
Datum a čas obhajoby: | 16.09.2019 09:00 |
Datum odevzdání elektronické podoby: | 19.07.2019 |
Datum odevzdání tištěné podoby: | 19.07.2019 |
Datum proběhlé obhajoby: | 16.09.2019 |
Oponenti: | Martin Flusser |
Konzultanti: | prof. RNDr. Ing. Martin Holeňa, CSc. |
Zásady pro vypracování |
The goal of this thesis is to survey the possibilities of application of
semi-supervised learning in sequence labeling using deep learning. The student shall: - study state of the art of semi-supervised deep learning - study possibilities of using semi-supervised deep learning for sequence labeling - propose a suitable semi-supervised deep learning strategy for sequence labeling in a field of video frame sequence processing or natural language processing - implement the proposed models - evaluate the models in terms of performance increase and learning time on a real-world sequential video or text data |
Seznam odborné literatury |
O. Chapelle, B. Schölkopf, and A. Zien. Semi-Supervised Learning. MIT
Press, Cambridge, Massachusetts, 2010. I. Goodfellow, Y. Bengio, and A. Courville. Deep Learning. MIT Press, 2016. http://www.deeplearningbook.org A. M. Dai and Q. V. Le. Semi-supervised sequence learning. In Advances in Neural Information Processing Systems 28, pages 3079-3087. Curran Associates, Inc., 2015. A. Oliver, A. Odena, C. Raffel, E. D. Cubuk, and I. Goodfellow. Realistic evaluation of semi-supervised learning algorithms. In NeurIPS, 2018. |
Předběžná náplň práce |
Deep learning has shown good performance in a variety of different
applications, especially in sequence modeling. Sequence models have garnered a lot of attention because most of the data in the current world are in the form of sequences. Such sequences might be formed in many different areas, from natural language processing through speech recognition and audio processing to video frame sequence processing. In the areas where a hidden, complex problem structure has to be extracted from the data, deep learning usually outperforms traditional statistical approaches. One disadvantage of deep learning is that it requires a large amount of data to be labeled and entirely available during training. Annotation of such datasets is usually slow and expensive. Semi-supervised learning could mitigate this problem by using cheaper unlabeled data together with the labeled data. It has found its application in several deep learning architectures. However, usage of semi-supervised deep learning for sequence modeling is very limited. |