Currently, a schema is a map from string-valued relation names, to a list of string-valued domain names:
typedef map<string, vector<string>> T_schema;
We will need to generalize this so that each relation is also associated with an output type (an output distribution).
There are a number of places in hirm.hh that explicitly construct a BetaBernoulli distribution. These will need to be modified, to construct the proper distribution object according to the schema.
Currently, a schema is a map from string-valued relation names, to a list of string-valued domain names:
We will need to generalize this so that each relation is also associated with an output type (an output distribution).
There are a number of places in
hirm.hh
that explicitly construct aBetaBernoulli
distribution. These will need to be modified, to construct the proper distribution object according to the schema.