v0.2.6

v0.2.6 (2019-10-10)

Diff since v0.2.5

Closed issues:

• Problem with show method for DenseEuclideanMetric (#105)
• Cannot use HMC and PG together in .2.4 (#95)
• Tempering with multiplication of momentum leap frog step (#52)
• Identify divergent transition (#50)
• Samples during adaption phase (#35)
• Support mutable and immutable array types (#18)

Merged pull requests:

• Fix effective depth (#111) (xukai92)
• Add kw to drop warmup samples (#108) (xukai92)
• Improve dense metric show (#107) (xukai92)
• Fix show(::DenseEuclidianMetrix) (#106) (cpfiffer)
• Implement tempering in leapfrog integrator (#96) (xukai92)

v0.2.5

v0.2.5 (2019-09-16)

Diff since v0.2.4

Closed issues:

• Initial mass matrix is reset to identity when n_adapts>0 (#98)

Merged pull requests:

• Init. adaptor via metric (#100) (xukai92)

v0.2.4

v0.2.4 (2019-09-03)

Diff since v0.2.3

Closed issues:

• More unit tests for HMC building blocks (#84)
• Possible (algorithmic) speedups (#32)
• Create a synthetic multivariate Gaussian test for pre-conditioning adaption (#25)
• Testing against CuArrays (#12)

Merged pull requests:

• Build interface for Turing (#93) (xukai92)

v0.2.3

v0.2.3 (2019-08-19)

Diff since v0.2.2

Closed issues:

• How to convert the output into a MCMCChains Chains (#89)

Merged pull requests:

• Update no-U-turn criteria and added some unit tests (#91) (xukai92)
• Working with CuArrays (#90) (xukai92)

v0.2.2

v0.2.2 (2019-07-26)

Diff since v0.2.1

Merged pull requests:

• Improve code: naming and numerical stability (#86) (xukai92)
• Fix mass matrix adaption resize function (#85) (xukai92)

v0.2.1

v0.2.1 (2019-07-25)

Diff since v0.2.0

Closed issues:

• Return more information for each step (#59)
• Implement Base.show for all user side types (#36)
• RFC Adapter types (#28)
• More robust and modular way of detecting divergence (#16)
• Adding noise to step size? (#10)

Merged pull requests:

• Bugfix for metric renew (#83) (xukai92)

v0.2.0

v0.2.0 (2019-07-23)

Diff since v0.1.9

Closed issues:

• Get more comments and docstrings (#75)
• Add more informative error messages. (#71)
• HMC: Multinomial Sampling Instead of Slice Sampling (#62)
• Update gradient function to return both value and gradient (#56)
• Simplifying Metric types by using AbstractPDMat (#34)
• RFC find_good_eps (#27)
• DynamicHMC is (~5 times) faster than Turing's HMC implementation (#26)

Merged pull requests:

• Towards v0.2.0 (#81) (xukai92)

v0.1.9

v0.1.9 (2019-07-05)

Diff since v0.1.8

Merged pull requests:

• Fix RFC typo (#78) (xukai92)
• Fixed bug preventing StanHMCAdaptor from adapting (#77) (chriselrod)
• Improving the performance and generality of adaptation (#74) (mohamed82008)

v0.1.8

v0.1.8 (2019-06-06)

Diff since v0.1.7

Merged pull requests:

• Bugfix for pre-conditioner dimensionality. (#72) (yebai)

v0.1.7

v0.1.7 (2019-05-31)

Diff since v0.1.6

Closed issues:

• Consider introduce the PhasePoint type (#17)

Merged pull requests:

• Fix callback functions without RNG for transition (#73) (yebai)
• Move rejection to MH (#68) (xukai92)