Permalink
Browse files

Moving explanation of concepts to extra file.

  • Loading branch information...
1 parent f437387 commit 1370ecd224830e3847252a80a63087b7bc5928fe Andreas Stuhlmueller committed Mar 29, 2011
Showing with 29 additions and 30 deletions.
  1. +0 −30 README.md
  2. +29 −0 docs/concepts.md
View
@@ -49,33 +49,3 @@ You should see the following output:
((#f #t) . 0.3333333333333333)
((#t #f) . 0.3333333333333333)
((#t #t) . 0.3333333333333333)
-
-## Concepts
-
-* _thunk_
-
- A procedure that takes no arguments.
-
-* _cc-cps-thunk_
-
- A thunk in closure-converted continuation-passing style. Whenever such a procedure gets to a random choice, it returns the current continuation, which is in hashable form due to cc.
-
-* _return-thunk_
-
- A cc-cps-thunk that has also undergone app conversion: All applications return to the top-level and pass function, continuation, and arguments back (all of which are hashable due to cc).
-
-* _graph_
-
- An acyclic graph for a probabilistic program, with one node for each random choice (as identified by its continuation and support). Corresponds to a system of linear equations.
-
-* _polygraph_
-
- A graph for a probabilistic program that makes subproblems explicit. For each subproblem, there is a parentless node, and references to the marginal probabilities of a subproblem are possible. Cyclic dependencies are only introduced via such references. Corresponds to a system of polynomial equations.
-
-* _polymap_
-
- A summary graph of the dependency structure of a polygraph. Contains a node for each subproblem (root node). Whenever a subproblem A references another subproblem B, there is a link from A to B in the polymap. In general, this graph is not acyclic.
-
-* _components_
-
- Clustering the strongly connected components of a polymap results in the acyclic components graph. Each component corresponds to a (linear or polynomial) problem that can be solved independently given the referenced parent parameters. By solving the components in topological order, the marginal distribution of the overall inference problem can be computed.
View
@@ -0,0 +1,29 @@
+# Concepts
+
+* _thunk_
+
+ A procedure that takes no arguments.
+
+* _cc-cps-thunk_
+
+ A thunk in closure-converted continuation-passing style. Whenever such a procedure gets to a random choice, it returns the current continuation, which is in hashable form due to cc.
+
+* _return-thunk_
+
+ A cc-cps-thunk that has also undergone app conversion: All applications return to the top-level and pass function, continuation, and arguments back (all of which are hashable due to cc).
+
+* _graph_
+
+ An acyclic graph for a probabilistic program, with one node for each random choice (as identified by its continuation and support). Corresponds to a system of linear equations.
+
+* _polygraph_
+
+ A graph for a probabilistic program that makes subproblems explicit. For each subproblem, there is a parentless node, and references to the marginal probabilities of a subproblem are possible. Cyclic dependencies are only introduced via such references. Corresponds to a system of polynomial equations.
+
+* _polymap_
+
+ A summary graph of the dependency structure of a polygraph. Contains a node for each subproblem (root node). Whenever a subproblem A references another subproblem B, there is a link from A to B in the polymap. In general, this graph is not acyclic.
+
+* _components_
+
+ Clustering the strongly connected components of a polymap results in the acyclic components graph. Each component corresponds to a (linear or polynomial) problem that can be solved independently given the referenced parent parameters. By solving the components in topological order, the marginal distribution of the overall inference problem can be computed.

0 comments on commit 1370ecd

Please sign in to comment.