Självlärande klassificering
(1971) In MSc ThesesDepartment of Automatic Control
- Abstract
- Consider a system recieving signals from a signal-source. The signals describe different input situations. The purpose of the system is to classify the current situation. This can be obtained by means of recursive algorithms. The problem of finding such algorithms can be approached in many different ways. In this report we will assume a minimum of apriori information about the signals and the classes they belong to. This approach is often called learning without a teacher or self-learning. Different self-learning algorithms are discussed and simulated on a computer. The conditions making an algorithm converge or not are treated. A theory for the convergence of an algorithm is presented. This theory is showed to be consistent with the... (More)
- Consider a system recieving signals from a signal-source. The signals describe different input situations. The purpose of the system is to classify the current situation. This can be obtained by means of recursive algorithms. The problem of finding such algorithms can be approached in many different ways. In this report we will assume a minimum of apriori information about the signals and the classes they belong to. This approach is often called learning without a teacher or self-learning. Different self-learning algorithms are discussed and simulated on a computer. The conditions making an algorithm converge or not are treated. A theory for the convergence of an algorithm is presented. This theory is showed to be consistent with the simulations. Finally another aspect is regarded implying no demand for stationarity of the classes. The algorithms are adapted to this case and found to work satisfactory. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/8850662
- author
- Bosrup, Lennart and Gustavi, Jan-Olof
- supervisor
- organization
- year
- 1971
- type
- H3 - Professional qualifications (4 Years - )
- subject
- publication/series
- MSc Theses
- report number
- TFRT-5097
- ISSN
- 0346-5500
- language
- Swedish
- id
- 8850662
- date added to LUP
- 2016-03-29 16:07:31
- date last changed
- 2016-03-29 16:07:31
@misc{8850662, abstract = {{Consider a system recieving signals from a signal-source. The signals describe different input situations. The purpose of the system is to classify the current situation. This can be obtained by means of recursive algorithms. The problem of finding such algorithms can be approached in many different ways. In this report we will assume a minimum of apriori information about the signals and the classes they belong to. This approach is often called learning without a teacher or self-learning. Different self-learning algorithms are discussed and simulated on a computer. The conditions making an algorithm converge or not are treated. A theory for the convergence of an algorithm is presented. This theory is showed to be consistent with the simulations. Finally another aspect is regarded implying no demand for stationarity of the classes. The algorithms are adapted to this case and found to work satisfactory.}}, author = {{Bosrup, Lennart and Gustavi, Jan-Olof}}, issn = {{0346-5500}}, language = {{swe}}, note = {{Student Paper}}, series = {{MSc Theses}}, title = {{Självlärande klassificering}}, year = {{1971}}, }