Package pycv :: Package cs :: Package ml :: Package cla :: Module dataset :: Class WeightedCDataset

Class WeightedCDataset

ml.Dataset --+    
             |    
  cla.CDataset --+
                 |
                WeightedCDataset

Known Subclasses:: ScoredWCDataset

A CDataset with weights attached to the samples.

Instance Methods

[hide private]

__init__(self, separated_input_data, separated_weights=None, approx=-1)

source code

new(self)
Make a copy of WeightedCDataset with a copy of weights and a reference to input_data.

source code

initialize_weights(self, value=-1)
Initialize weights if 'weights' is None

source code

set_weights(self, value=1)
Set all weight values to 'value', initialize if necessary

source code

get_twpc(self)
Get total weights per class.

source code

get_skewness(self)

source code

get_tw(self)
Get total weights of all class(es).

source code

normalize_weights(self)
Normalize the weights so that the sum of all weights is equal to 1.

source code

scale_weights(self, i, thelambda)
Scale the weights of class i by thelambda.

source code

dofilter(self, j, filterarray)
Filter away some samples in class j.

source code

concat(self, j, input_data, weights=None)
Concatenate a few samples to class j.

source code

compute_Stats2m(self)
Compute moments up to 2nd-order.

source code

_get_masks(self, j)
Get the masks for instances of class j if approximation is needed.

source code

compute_Stats2e(self)
Compute per-element moments up to 2nd-order.

source code

compute_Stats2(self)
Compute (total weights, means, and covariances).

source code

Method Details

[hide private]

init(self, separated_input_data, separated_weights=None, approx=-1)
(Constructor)

source code

Overrides: cla.CDataset.__init__

dofilter(self, j, filterarray)

source code

Filter away some samples in class j.

Input:

j: class j filterarray: a 'bool' numpy.array of size self.nspc[j]

Output:

fr: filtering rate, the new number of samples of class j divided: by the old number of samples of class j. The WeightedCDataset is updated.

Returns:

fr: filtering rate, the new number of samples of class j divided: by the old number of samples of class j. The class is updated.

Overrides: cla.CDataset.dofilter

concat(self, j, input_data, weights=None)

source code

Concatenate a few samples to class j.

Input:
    j: class j
    input_data: an array of samples
    weights: the additional weights of the samples
        if weights is None xor self.weights is None:
            raise an error
        else:
            concatenate the new weights to the old weights list
Output:
    The WeightedCDataset is updated.

Overrides: cla.CDataset.concat

compute_Stats2m(self)

source code

Compute moments up to 2nd-order. The results are stored as a single Stats2m.

Output:: B: a Stats2m representing the obtained statistics

compute_Stats2e(self)

source code

Compute per-element moments up to 2nd-order. The results are stored as a single Stats2e.

Output:: B: a Stats2e representing the obtained statistics

compute_Stats2(self)

source code

Compute (total weights, means, and covariances). The results are stored as a single Stats2.

Output:: B: a Stats2 representing the obtained statistics

Class WeightedCDataset

__init__(self, separated_input_data, separated_weights=None, approx=-1) (Constructor)

dofilter(self, j, filterarray)

concat(self, j, input_data, weights=None)

compute_Stats2m(self)

compute_Stats2e(self)

compute_Stats2(self)

init(self, separated_input_data, separated_weights=None, approx=-1)
(Constructor)