Roadmap for v0.3

[yebai]

• NUTS implementation NUTS implementation #188
• HMC: Transforms of ϵ for each variable HMC: add pre-conditioning (mass) matrix #67 (replace with introducing mass matrix)
• Re-define ~: macro => function Re-define ~: macro => function #173 (postponed to 0.4)
• Finish: Sampler (internal) interface design Sampler (internal) interface design #107
• Substantially improve performance of HMC and Gibbs Automated benchmarking for Gibbs, HMC #7
  • benchmark based on NUTS only
  • translate more Stan models into Turing (for NUTS): https://github.com/stan-dev/example-models
  • vectorising likelihood computations vectorising likelihood computations #117 Vectorization of assume #255
  • Time waste for constructing distributions in Distributions.jl Time waste for constructing distributions in Distributions.jl #166
• Remove obsolete randoc, randc? remove obsolete randoc, randc? #156
• Maximum a posteriori (MAP) inference. Maximum a posteriori (MAP) inference. #72
• Compositional MCMC: partial refreshment of the momentum variable Compositional MCMC: partial refreshment of the momentum variable (lightNUTS) #189
• Allowing user to change the compiler between safe mode and fast mode
• Support truncated distribution. Support truncated distribution. #87
• Refactoring code: Unify VarInfo, Trace, TaskLocalStorage Refactoring code: Unify VarInfo, Trace, TaskStorage, Chain. #96
• Refactoring code: Better gradient interface Better gradient interface #97
• Support sampling from model in model #86