Algorithm-Burg

 view release on metacpan or  search on metacpan

README  view on Meta::CPAN

VERSION

    version 0.001

SYNOPSIS

DESCRIPTION

    The Algorithm::Burg module uses the Burg method to fit an
    autoregressive (AR) model to the input data by minimizing (least
    squares) the forward and backward prediction errors while constraining
    the AR parameters to satisfy the Levinson-Durbin recursion.

    DISCLAIMER: This is work in progress! The code is buggy and the
    interface is subject to change.

ATTRIBUTES

 coefficients

    AR model polynomial coefficients computed by the train method.

lib/Algorithm/Burg.pm  view on Meta::CPAN

=head1 VERSION

version 0.001

=head1 SYNOPSIS

=head1 DESCRIPTION

The L<Algorithm::Burg> module uses the Burg method to fit an autoregressive (AR)
model to the input data by minimizing (least squares) the forward and backward
prediction errors while constraining the AR parameters to satisfy the
Levinson-Durbin recursion.

B<DISCLAIMER: This is work in progress! The code is buggy and the interface is subject to change.>

=head1 ATTRIBUTES

=head2 coefficients

AR model polynomial coefficients computed by the C<train> method.

t/01-coeffs.t  view on Meta::CPAN

    $burg->train([
        map {
              1.0 * cos($_ * 0.01) + 0.75 * cos($_ * 0.03)
            + 0.5 * cos($_ * 0.05) + 0.25 * cos($_ * 0.11)
        } 0 .. 127
    ])
};

is($#coeff1, $#coeff2, 'order');

my $error = 0.0;
my $epsilon = 1e-9;
for my $i (0 .. $#coeff1) {
    my $delta = $coeff1[$i] - $coeff2[$i];
    $error += $delta * $delta;
}

ok($error < $epsilon, "error < epsilon ($error)");

t/02-predict.t  view on Meta::CPAN

$burg->train([ @whole_series[0 .. 64] ]);

my @original = @whole_series[64 .. $#whole_series];
my @predicted = @{ $burg->predict(64) };

is($#original, $#predicted, 'cardinality');

# use Data::Dumper;
# diag Dumper(\@original, \@predicted);

my $error = 0.0;
my $epsilon = 3.0;
for my $i (0 .. $#original) {
    my $delta = $original[$i] - $predicted[$i];
    $error += $delta * $delta;
}

ok($error < $epsilon, "error < epsilon ($error)");