Graphical Model Profiling Tools

This tool converts stan model file (left) into variable dependencies as graph structure, and visualize the graph (right).

Getting Start

Put your stan files in code/, and git clone this repo such that your working directory contains both code/ and Profile_Stan/.

mkdir outputs
mkdir outputs/probgraph

and run

python model_parser_wrapper.py

To visualize the graph, run

python to_graphviz.py ../outputs/probgraph/[the graph to visualize].probgraph

and paste the results to http://www.webgraphviz.com.

Probabilistic Graph Data Structure

parse_model.py stores probability graph ending with .probgraph. It streams graph into a json file. The data structure is:

• {node: [([parents], dependency_type)]} where
• dependency_type:
  • indexing,
  • simple/complex computation,
  • discrete/continuous distribution

CAVEAT

1. variables are dependent on the index variables
2. real beta[3]; there are 3 betas here, create "3" as a "variable"; only keep the largest integer
3. ignore the arithmetic operations within []
4. examples: eps[2:nyear] = eps2[1:nyear - 1], or N_est[t + 1] = N_est[t] * lambda[t];
5. a[b[i],c[i]], generate i->b, i->c, b->a, c->a (ignore i->a)
6. verified 14 models out of 69 models in BPA
7. does not work for one statement with {} in one line
8. does not support /* or */ appearing after statements
9. simply deal with functions{} (e.g. Ch.07/cjs_add.stan) by adding a connection between returned value and arguments
10. the dependency belongs to complex
11. (a more sophisticated way to solve this is to build a graph for each functions, and map variables accordingly whenever the function is called)
12. works for 69 model files in BPA, verified 17 files
13. variables assigned within if statement, are dependent on the variables in if conditions. Dependency type is basic.
14. variables declared within for loops, are dependent on the loop indices; the dependency is indexing
15. we add indexing dependency for one variable at a time. For example (['a', 'b'], 'indexing') should be converted to (['a'], 'indexing'), (['b'], 'indexing')
16. does not work for for statement without {} followed by more than one statement, example: 28 for (n in 1:nmax) 29 if (n < nmin) 30 lp_parts[n] <- log(1.0 / nmax) + negative_infinity(); // Zero probability 31 else 32 lp_parts[n] <- log(1.0 / nmax) + binomial_log(k, n, theta);
17. does not work for if/for statements, whoes { is in a new line
18. when calling increment_log_prob(), construct a 'target' variable to hold the value
19. does not work if for/if statement put condition and statement together..

Emma Wang Originally on Feb 24th, 2017, Updated on March 21th, 2017.