v0.1.0a54

• Use yail for many functions from nanshe.util.iters. #425
• Optimization and drop bad doctest in nanshe.util.iters. #424
• Use toolz for some functions in nanshe.util.iters. #423
• Fix a broken doctest in nanshe.util.iters. #422
• Use xnumpy for nanshe.util.xnumpy expand view functions. #420
• Use npctypes for nanshe.util.xnumpy type functions. #419 #421
• Drop use of nomkl. #418
• Fix documentation in nanshe.box.spams_sandbox. #417
• Refactor nanshe.box.spams_sandbox to use npctypes. #416
• Unvendor dependencies and make external requirements. #415
• Remove unneeded dependencies from the build. #414
• Relicense as BSD 3-Clause. #411 #412
• Use vendored kenjutsu for nanshe.util.iters slice operations. #410
• Unpin sphinx as recent version have needed features. #407
• Update line coverage exclusions (vendored, drop ilastik). #406
• Rename vendored directory. #405
• Adjust nose ignore to exclude vendored. #404
• Disk usage optimizations for Wercker. #403
• Prep for vendoring dependencies. #402
• Tidying of vendored dependencies. #401
• Have git ignore .wercker/. #400
• Upgrade to versioneer version 0.16. #399
• Pin to conda-build version 1.* in CI. #398
• Use vendored metawrap in place of nanshe.util.wrappers. #397
• Provide CI testing with/without GPL requirements. #396
• Drop optional GPL requirements from dependencies. #396
• Drop PyQt utility functions (will be packaged separately). #395
• Remove Qt/PyQt dependency. #395
• Drop dependency on volumina. #394
• Remove nanshe.viewer (will be packaged separately). #394
• Fixes to Travis CI to handle SGE changes. #393
• Rewrite HDF5 path handling from scratch. #392
• Drop vendored copy of pathHelpers. #392
• Suppress TRACE logging by default. #353

v0.1.0a53: Some minor improvements.

• Refactor wavelet_denoising into smaller functions.
• Fix some subtle Python 3.x issues btwn NumPy and multiprocessing.
• Prepare for use of SPAMS on Python 3.x.

v0.1.0a52: Update to HDF5 1.8.17[.*]

v0.1.0a52: Update to HDF5 1.8.17[.*].

v0.1.0a51: Use conda-forge and other minor changes.

• Start using conda-forge as dependencies basically live there.
• Some small Python 2/3 fixes with regards to map.
• Skip setup.py when running nose tests.

v0.1.0a50: Scikit-image fixes and other minor changes.

• Add fixes to region properties for scikit-image 0.12.0.
• Update SPAMS warning to be a bit more clear.
• Minor memory optimization for registration.

v0.1.0a49: Formal CI and deployment for Python 3.

• Adds Travis CI support for Python 3 with some constraints.
• Fixes a testing hang with some of the SPAMS sandboxing code.

v0.1.0a48: Cleanup syntax so as to support Python 2/3.

• Changed syntax to support Python 2/3. (CI in the works)
• Various CI improvements.
• Minor code style improvements.

v0.1.0a47: Allow use of dictionary learning from scikit-learn too.

• Add support for using scikit-learn in dictionary learning or SPAMS.
• Provide unit tests for using scikit-learn for dictionary learning.
• Add optional method-agnostic number of components argument in dictionary learning.
• Change learner regression test parameters to be more realistic.
• Use fewer blocks in 2D learner regression tests to speed them up.
• Add Travis CI SGE/DRMAA testing back (now that tests are fast enough).
• Correct comparison of max intensity points with centroids.
• Fix some dictionary tests that were missing the number of basis images.
• Log number of accepted neurons in parallel learner.
• Fix various formatting issues.

v0.1.0a46: Require `nomkl`.

• During conda builds and installs explicitly require nomkl.

v0.1.0a45: Ensure MKL is not used.

• Avoid the error prone testing step in package building.
• Fix CI to not use MKL.