release-2.0

• Complete rewrite of the Makefiles to simplify them and make them
  compatible with Matlab 2014a (which broke a lot of things). The
  Makefiles no longer attempt to detect which compiler MEX uses, so
  you'll have to set it manually if you use a nonstandard one (e.g.,
  'make CXX=g++-4.7.1 CC=gcc-4.7.1'). This has been tested on 64-bit
  Linux and Mac OS.
• When constructing a braid object from data, the braid constructor
  now uses a C++ MEX file (written by Marko Budisic) giving a big
  speedup for large datasets. Support for parallelization on multiple
  cores is included. The global variable BRAIDLAB_threads determines
  the number of threads used.
• The braid constructor can make knot representatives, e.g.,
  b=braid('8_3') returns a braid representative for the third
  8-crossing knot.
• The private method braid.loopsigma's C++ helper functions (used by
  braid*loop) can now handle int32, int64, and single, in addition to
  double. This makes checking braid equality and triviality much
  faster, unless overflow occurs and VPI has to be used.
• The private method braid.loopsigma's C++ helper functions can also
  handle VPI (Variable Precision Integers) by using GMP, the GNU
  MultiPrecision library. If you don't have GMP installed, compile
  with option "BRAIDLAB_USE_GMP=0". However, this will lead to a
  massive slowdown when working with VPI types (Variable Precision
  Integers).
• New methods in the loop class: loop.components, loop.plot
  'Components' option, loop.getgraph. These are not yet documented in
  the guide, as they are an advanced feature, though they are
  described in the help text to the functions.
• braid.burau and braid.alexpoly can use the symbolic toolbox.
  alexpoly no longer centers the polynomial by default, since
  centering cannot always be used with a laurpoly object, and can
  never be used for integral types. Use option 'centered'.
• braid.mtimes acting on loops has an optional output argument that
  records the signs of the pos/neg operators in the update rules
  (loopsigma). This allows reconstruction of the matrix for the
  effective linear action, as well as checking for limit cycles for
  the action.
• New methods in the braid class: braid.linact converts the optional
  output from braid.mtimes (see above) to a matrix. braid.cycle looks
  for a limit cycle in the braid action on loops. braid.cyclemat uses
  braid.cycle to return a matrix corresponding to this limit cycle.
  The largest eigenvalue of this matrix, normalized by the period,
  gives the Perron root of the largest pseudo-Anosov component(s).
  This is not documented in the guide yet, because it's still a bit
  experimental. In particular it sometimes fails to find the limit
  cycle (issue #52).
• Many other small bugfixes and improvements.

Several improvements to the method braid.entropy:

• braid.entropy has a C MEX helper function and is much faster. It no
  longer checks if the braid is trivial, which avoids overflow and
  speeds up the function.
• braid.entropy's convergence has been greatly improved (see next item
  for more details). With enough iterations, machine precision can
  usually be achieved.
• braid.entropy's optional second return argument has changed from a
  list of iterates to the final generalized eigenvector. The reason
  is that the internal algorithm has changed: the Dynnikov coordinates
  get renormalized at every step, much like the power iteration method
  for finding the largest eigenvalue of a matrix. A consequence of
  this is that it is no longer natural to store the iterates. This
  shouldn't affect things very much, since this optional argument was
  not used widely.

release-1.0.5

• Constructors for Rupert Venzke's psi-family of low-entropy braids.
• Fix bug involving single-trajectory dataset.
• Plot empty braids (why not...).
• braid.compact no longer checks for trivial braid, since this can
  take a lot of time (more than the compact itself).
• Unit test for entropy added to testsuite.
• Lighter distribution (just compiled files).
• A few more small bugfixes.

release-1.0.4

• The databraid class records crossing times of a dataset.
• Added an example section on taffy pullers to braidlab guide.
• bugfix in testsuite.

release-1.0.3

• The braidlab guide has been updated and expanded, and now has a
  table of contents and a detailed index.
• New methods in braid class: 'alexpoly' returns the Alexander
  polynomial. 'burau' can now construct a matrix of Laurent
  polynomials (wavelet toolbox required). 'tensor' is the tensor
  product of braids.
• Big speedup in creating braids from data, due to improved safety
  checks on the particle trajectories.

release-1.0.2

• loopTest.m in testsuite.
• Clean up files. Eliminate subrepos.
• First public release.

release-1.0.1

• Check for overflow of loop coordinates.
• Several bugfixes.
• Improved support for VPI (Variable Precision Integers).

release-1.0

First release of braidlab.