Marsyas
0.6.0-alpha
|
Calculate Alignment from a SimilarityMatrix at input. More...
#include <DTW.h>
Inherits MarSystem.
Public Member Functions | |
MarSystem * | clone () const |
DTW (std::string name) | |
DTW (const DTW &a) | |
void | myProcess (realvec &in, realvec &out) |
Processes data. | |
~DTW () |
Calculate Alignment from a SimilarityMatrix at input.
as input format is expected: in combination with the sizes control - which has to be of size n where n is the number of sequences, and each element of the control has to be the number of vector-elements in the nth sequence - the first 1/n part of each row containing the elements of the vectors of the first sequence the second 1/n part containing the elements of the vectors of the second, and so on up to nth part containing the elements of the vectors of the nth sequence.
Controls:
CAUTION: since the number of elements of the sizes control is assumed to be the number of seqences to be compared, the dimensionality of the vectors is assumed to be "numberOfInputRows(i.e. Observations) divided by numberOfSequences(i.e. length of the sizes control)" since all vectors have to be of the same dimensionality. e.g. the standard case of two sequences would need to have the sizes control of length 2 with sizes(0) being the length of the fist sequence, sizes(1) being the length of the second; and each input observation (i.e. row) containing two input vectors: the first half of the row containing the elements of the vectors of the first sequence the second half of the row containing the elements of the vectors of the second sequence
Output size: When the control mode is set as normal, output size should be inObservations+inSamples. When the contorl mode is set as OnePass, output size should be 3*inSamples.