1.3 What is this Thesis about?
1.3.1 Thesis Questions
This thesis studies one specific ensemble learning technique, Negative Correlation (NC)
Learning. NC has shown several empirical successes [18, 85, 86, 93, 152], often outperforming
other ensemble learning methods on a variety of tasks, but as yet has had no explanations
offered as to why it performs well when it does. As such, the primary thesis question is
“Why and how does NC Learning succeed as an ensemble method?”. NC has a problemdependent
parameter, λ, which determines the amount of diversity among the ensemble
members. This parameter can determine how well NC performs, yet currently there is little
information on how to set it. A secondary thesis question is therefore “How can we provide
guidance for setting λ in the NC learning algorithm?”.
1.3 What is this Thesis about?1.3.1 Thesis QuestionsThis thesis studies one specific ensemble learning technique, Negative Correlation (NC)Learning. NC has shown several empirical successes [18, 85, 86, 93, 152], often outperformingother ensemble learning methods on a variety of tasks, but as yet has had no explanationsoffered as to why it performs well when it does. As such, the primary thesis question is“Why and how does NC Learning succeed as an ensemble method?”. NC has a problemdependentparameter, λ, which determines the amount of diversity among the ensemblemembers. This parameter can determine how well NC performs, yet currently there is littleinformation on how to set it. A secondary thesis question is therefore “How can we provideguidance for setting λ in the NC learning algorithm?”.
การแปล กรุณารอสักครู่..