a=algorithm(hyperParam)
An algorithm with given hyperparamters is constructed.
All algorithms will inherit these hyperparamters resp. option settings.
Hyperparameters/Option Settings (with defaults):
training_time=0 -- cputime needed for training is taken
trained=0 -- true, if a model has been learnt
do_not_retrain=0 -- determines if the algorithm has to
be retrained (useful for param)
use_prev_train=0 -- determines if previous model shall
be used (e.g. param objects don't need retraining)
do_not_evaluate_training_error=0 -- selfexplanatory (speed up computation)
use_signed_output=1 -- true if sign after output
shall be taken
verbosity=1 -- verbosity level
is_data=0 -- true if object is considered as data
alias=[] -- alternative names for this
object
-- accessing members, e.g can do:
a=svm;
a.alias={'p1','C','p2','alpha'};
a.p1=5; a.p2=ones(1,10);
deferred = [];
Methods:
train,test -- selfexplanatory