One kind constructs a set of base learners called homogeneous base learners with a single base learning algorithm; the other kind produces base learners by adopting multiple learning algorithms, which are called heterogeneous learners.
After generating a set of base learners, the ensemble method will then be formed by combining the base models or a subset of them based on defined criteria or algorithm to form a generalised prediction model.
The target model is a weighted ensemble of a set of basic learners.
Bagging trains a set of base learners each from a different dataset with the same size as the training dataset obtained by subsampling the training dataset with replacement.
Other machine learning algorithms may be considered (for example, Breiman, 2001), as well as boosting to improve on a set of weak learners (Bühlmann, 2007).
To reduce the model complexity and training time, we propose a novel sparse Bayesian MIML learning model, which applies a RVM-like algorithm to obtain a sparse weighted combination for a set of basic learners.
Since the cost of calculating the optimal KGP is very high, we randomly set them Q times to obtain a set of different learners and then apply a weighted ensemble procedure as follows: (7) f ens (x ) = ∑ i = 1 Q f i (x ).
There are four main phases: (1) goal map building and generation of a set of parts, (2) learner map building, (3) map diagnosis and goal map modification, and (4) learner map modification.
In the next section, the framework of KB map is explained as a flow of map building composed of (1) goal map building and generation of a set of parts, (2) learner map building, (3) map diagnosis and goal map modification, and (4) learner map modification.
In order to conduct this initial grouping process, we apply a clustering technique that brings a seed set of learners into an initial set of CoPs.
