v0.3.2

MCMCTempering v0.3.2

Diff since v0.3.1

Closed issues:

• Leverage existing ways in TuringLang (if they exist) / Julia ecosystem to approximate K in the case of Simulated Tempering (#3)
• Is there any reason to return all of the PT chains, or should only the one where beta = 1 be returned in all cases? (#20)
• Solve the problem of sampler specific types obscuring access to the VarInfo (#63)
• Figure out a consistent way to return swap history and other chains if desired (#123)
• Extend the suite of tests to match level of coverage in Turing (#124)
• Support multiple samplers (#148)
• Support passing a chain type and returning in line with it from tempered_sample (#149)

Merged pull requests:

• Renaming of some fields (#142) (@torfjelde)
• Implementing and correcting the base set of swap strategies, plus wrapping swap moves as part of steps (#143) (@HarrisonWilde)
• Introduction of compositions and products of samplers and models (#151) (@torfjelde)
• Introduction of SwapSampler + make TemperedSampler a fancy version of CompositionSampler (#152) (@torfjelde)
• Documentation (#153) (@torfjelde)
• Use process_to_chain in roundtrips instead (#154) (@torfjelde)
• Improvements to state_from and more extensive docs (#155) (@torfjelde)