NEWS.md

In IncDTW: Incremental Calculation of Dynamic Time Warping

Version 1.1.4.04

other changes

• Bug fix in unit tests

Version 1.1.4.03

other changes

• Updated references and citation information to link to the JSS paper DOI: 10.18637/jss.v099.i09

Version 1.1.4.02

Bug Fixes

• missing documentation for dist_method of lowerbound()
• There was a bug in the test file test_cm.r. I replaced test_that() with expect_equal().

New Features

• ...

other changes

• changed weblink of ref in descrition of DBA()

Version 1.1.4.01

Bug Fixes

• there was a buggy test "Double Incremental Matrix EQUAL Scratch" in the test file "test_dtw.R". The error ocurred for some random seeds when the optimal warping path is not unique. The DTW distance measure and the GCM is always unique and correct, but the direction matrix and warping path are not. I set a seed, so that the error should not happen again in this test file.

New Features

• new function: lowerbound() and lowerbound_tube()
• ...

other changes

• I reduced the file size of the vignette "Theory and Applications for the RPackage IncDTW.pdf" by about 800Kb.
• ...

Version 1.1.3

Bug Fixes

• if parameter ws = Inf then R breaks. A new check function checks if ws = Inf and if TRUE, then ws is set to NULL, which is equivalent to the meaning of Inf.
• scale() returned NaN if the standrad deviation was 0. Now there is a check if the sd is smaller than the parameter threshold (default = 1e-5). If the sd is smaller than the threshold, then no scaling is performed, only shifting. Analogous it's implemented for min-max scaling. Also for rundtw().

New Features

• New vignette that replaces the old vignettes
• replaced the function norm() by scale(), same functionality. norm() still works, but prints a warning to be deprecated.
• replaced the arguments 'normalize' by 'scale' for the function rundtw(). See details of the function documentation.

other changes

• Changed the name of the data set "Walk" to "walk"
• For clarification I replaced 'norm' by 'scale' in the context of z-scaling and min-max-scaling (z-normalization and min-max-normalization). From now on the terminology should be clearer seperated from normalizing the DTW distance for the length of the time series.

Version 1.1.2

Bug Fixes

• fixed the case of calculating the cost matrix with cm() for univariate time series with a self defined distance function.
• export the 'insert' functions for simulate_timewarp()

New Features

• new branch of wrapper functions around the new S3 class 'planedtw': initialize_plane(), increment(), decrement(), reverse(), refresh(). This set of function should make it easier to navigate in the plane of possible fits, to increase the usability of the functions idtw2vec() dtw_partial() that are called behind the scenes. Also plot() and print() methods for the class 'planedtw'.
• an improved 'lot-mode' for rundtw() -- where the parameter 'C' is a list of time series -- helps to keep the allocated storage low
• new parameter '...' for the function cm() allows to pass further arguments
• new S3 class 'rundtw' for results of the function rundtw()
• print and summary methods for the S3 classes 'idtw', 'dba' and 'rundtw'
• plot method for the S3 class 'rundtw'
• new parameter 'return_QC' for rundtw() for easier plotting
• is.class() for all S3 classes in the package

other changes

• revised all examples in the help files
• renamed DBA() to dba(). DBA() is still available, but deprecated. A warning is printed.

Version 1.1.1

Bug Fixes

• fixed issue with initial best-sofar-value-in-window for rundtw(), too many unnecessary computations were completed within the first nh observations

New Features

• running z-normalization for the function rundtw(). Up to now only min-max-normalization was implemented. The parameter 'normalize' now has 3 possible values, the former TRUE and FALSE are still possible to pass. They will be translated internally to '01' and 'none' and a warning message is printed, saying that TRUE and FALSE is deprecated.
• lot-mode: ('list-of-timeseries'-mode) the parameter 'C' for the function rundtw() can now also be a list of time series. So rundtw() can search for the kNN of a query pattern Q in a list of time series of varying lengths.
• new entry in the result vector 'counter' of the function rundtw(): counter["completed"]

other changes

• add-on in the description of the data sets. That it's z-normalized.

Version 1.1.0

Bug Fixes

• add labels to result of dtw_dismat() and dtw_disvec()
• corrected assignment of ii and jj to Q and C in the description files

New Features

• new function: rundtw()
• new function: find_peaks()
• new feature: for simulate_timewarp(), the parameter preserve_length
• vector based (also incremental) implementation for existing cost matrix
• plot functions for DBA for multivariate time series

other changes

• change name of Vignette, the name visible online
• new Vignette, which is an extensive discussion of DTW, incremental DTW and sub sequence matching

Version 1.0.5

Bug Fixes

• normalized dtw in dtw() for multivar time series

New Features

• simulate_timewarp(): new function
• dtw2vec_cm() and idtw2vec_cm() : new functions included in dtw2vec() and idtw2vec()

IncDTW documentation built on March 18, 2022, 6:43 p.m.