commit f32cdeb070f72279b689d76ff3a842a0e0e0bdb7
parent 18eef87251dd0a4ca2f40c76cc0c06403b063f2c
Author: Jared Tobin <jared@jtobin.ca>
Date: Tue, 13 Oct 2015 23:18:20 +1300
Docs update.
Diffstat:
2 files changed, 16 insertions(+), 9 deletions(-)
diff --git a/README.md b/README.md
@@ -62,11 +62,11 @@ Installing is best done via
everything you might need (including GHC).
You'll want to use the [Stackage nightly
-resolver](https://www.stackage.org/nightly) until the next LTS version picks
-up these libraries.
-
-With that out of the way it's just a matter of
+resolver](https://www.stackage.org/nightly) for now, until the next LTS version
+is released. But with that out of the way it's just a matter of
```
$ stack install declarative
```
+
+See the test suite for some example usage.
diff --git a/declarative.cabal b/declarative.cabal
@@ -1,5 +1,5 @@
name: declarative
-version: 0.1.0.1
+version: 0.1.1
synopsis: DIY Markov Chains.
homepage: http://github.com/jtobin/declarative
license: MIT
@@ -10,12 +10,19 @@ category: Math
build-type: Simple
cabal-version: >=1.10
description:
- DIY Markov Chains.
+ This package presents a simple combinator language for Markov transition
+ operators that are useful in MCMC.
.
- Build composite Markov transition operators from existing ones for fun and
- profit.
+ Any transition operators sharing the same stationary distribution and obeying
+ the Markov and reversibility properties can be combined in a couple of ways,
+ such that the resulting operator preserves the stationary distribution and
+ desirable properties amenable for MCMC.
.
- A useful strategy is to hedge one's sampling risk by occasionally
+ We can deterministically concatenate operators end-to-end, or sample from
+ a collection of them according to some probability distribution. See
+ <http://www.stat.umn.edu/geyer/f05/8931/n1998.pdf Geyer, 2005> for details.
+ .
+ A useful strategy is to hedge one's 'sampling risk' by occasionally
interleaving a computationally-expensive transition (such as a gradient-based
algorithm like Hamiltonian Monte Carlo or NUTS) with cheap Metropolis
transitions.
