Share on email
Share on print
Share on whatsapp
Share on skype
Share on google
Share on facebook
Share on twitter

Groenewald, Van Huyssteen & Puttkammer 2007

Groenewald, Hendrik J., Gerhard B. Van Huyssteen, and Martin J. Puttkammer. 2007. “Evaluating wrapped progressive sampling for automatic algorithmic parameter optimisation.” International Conference Recent Advances in Natural Language Processing, RANLP:251-255.

  • Files
  • Keywords
  • Abstract
  • Languages

English: optimisation, machine learning, parameters, lemmatisation, hyphenation

Afrikaans: lemmatisering, masjienleer, optimering, parameters, woordafbreking


English: Determining the algorithmic parameter combinations that deliver the best performance in applications using machine learning algorithms is a very important part in the development process. Exhaustive searches are slow and  computationally expensive, which motivates the investigation  of more efficient methods of automatic algorithmic parameter  optimisation. Wrapped progressive sampling is one such a  method and is utilised in a tool named Paramsearch. An  alternative method for determining the sizes of the  progressive datasets used in the wrapped progressive  sampling procedure is proposed and implemented as  PSearch. PSearch and Paramsearch are evaluated and compared to an exhaustive search on the tasks of  lemmatisation and hyphenation in Afrikaans. Results indicate  that both PSearch and Paramsearch are generally more  efficient in terms of execution time and computational  resources than an exhaustive search. It is also shown that  PSearch delivers more accurate results than Paramsearch on  the tasks of lemmatisation and hyphenation in Afrikaans.

In: English

On: Afrikaans