NAME

Algorithm::LibLinear::ScalingParameter

SYNOPSIS

  use Algorithm::LibLinear::DataSet;
  use Algorithm::LibLinear::ScalingParameter;
  
  my $training_data = Algorithm::LibLinear::DataSet->load(fh => \*DATA);
  my $parameter = Algorithm::LibLinear::ScalingParameter->new(data_set => $training_data);
  my $scaled_training_data = $training_data->scale(parameter => $parameter);
  say $scaled_training_data->as_string;
  # 1 1:0.8541665 2:1 3:1 4:0.3396225 5:0.4474885 6:0 7:1 8:0.2900765 9:0 10:0.387097 12:1 13:0
  # -1 1:0.7916665 2:0 3:0.6666665 4:0.198113 5:1 6:0 7:1 8:0.6793895 9:0 10:0.2580645 12:0 13:1
  # 1 1:0.5833335 2:1 3:0.3333335 4:0.283019 5:0.308219 6:0 7:0 8:0.53435115 9:0 10:0.048387 11:0 12:0 13:1
  # ...
  
  __DATA__
  +1 1:0.708333 2:1 3:1 4:-0.320755 5:-0.105023 6:-1 7:1 8:-0.419847 9:-1 10:-0.225806 12:1 13:-1 
  -1 1:0.583333 2:-1 3:0.333333 4:-0.603774 5:1 6:-1 7:1 8:0.358779 9:-1 10:-0.483871 12:-1 13:1 
  +1 1:0.166667 2:1 3:-0.333333 4:-0.433962 5:-0.383562 6:-1 7:-1 8:0.0687023 9:-1 10:-0.903226 11:-1 12:-1 13:1 
  ...

DESCRIPTION

Algorithm::LibLinear::ScalingParameter contains configuration for feature scaling. The parameter is used by Algorithm::LibLinear::DataSet's scale method.

METHODS

new([data_set => $data_set] [, lower_bound => 0] [, upper_bound => 1.0] [, min_max_values => \@min_max_values])

Constructor. At least you have to specify either data_set or min_max_values. data_set is an instance of Algorithm::LibLinear::DataSet. min_max_values is an ArrayRef of 2-element ArrayRefs. Each ArrayRef should contain minimum and maximum value of corresponding feature.

load([fh => \*FH] [, filename => $path] [, string => $string])

Class method.

Restores parameter from LIBSVM/LIBLINEAR scaling parameter file.

as_string

Dumps scaling parameter as LIBSVM/LIBLINEAR format.

lower_bound

upper_bound

The lower/upper bound of scaled feature value.

save([filename => $path] [, fh => \*FH])

Writes scaling parameter into a file.