Output

Phillip's outputs are written in XML format.

Configure

All kinds of output from Phillip include the common header. This header contains the configure on inference, such as the version, the parameters, the flags and the outline of the knowledge base used.

• version Provides the version of Phillip.
• components Provides the name of components used. The attribute lhs corresponds to an enumerator of latent hypotheses set. The attribute ilp corresponds to a converter from latent hypotheses set into ILP problems. The attribute sol corresponds to an ILP solver.
• knowledge_base Provides the outline of the knowledge base used.
• params Provides the parameters used.

Latent hypotheses set

Using the option -p path_lhs_out=<PATH>, you can get the detail of the latent hypotheses set. Its elements are as follows:

• latent-hypotheses-set The root. name corresponds the name of the observation. time means the time which is spent on enumeration of a latent hypotheses set. timeout means whether the enumeration have been interrupted by timeout.
• nodes Contains the information of nodes in the latent hypotheses set. num means the number of nodes.
  • node Provides information of each node. index means the node's ID number. depth means the node's inference chain depth. The depth of an observed literal is 0. The depth of an equality or a required literal is -1. master means the index of the master hypernode, which is the head of the edge hypothesizing the node. This value of observed literal of equality or required literal is -1.
• axioms Provides the all axioms which have been used on enumeration of a latent hypotheses set. num means the number of axioms used.
  • axiom Provides information of each node. id means the axiom's ID number. name corresponds the axiom's name.
• edges Provides the all edges in a latent hypotheses set. num means the number of edges.
  • edge Provides information of each edge. id means the edge's ID number. type means the type of the edge. tail means the index of the hypernode which corresponds to the edge's tail. head means the index of the hypernode which corresponds to the edge's head. axiom means the ID number of axiom which used in this edge. If the edge does not use any axiom (e.g. an unification edge), this value is -1. conds means conditions which this edge needs to satisfy.
• substitutions Provides clusters of unifiable variables in a latent hypotheses set.
  • cluster Provides information of an unifiable variables cluster. id means the cluster's ID number.
• mutual_exclusive_nodes Provides mutual exclusions between two nodes. num means the number of mutual exclusive nodes pairs in the latent hypotheses set.
  • xor Provides information of mutual exclusion between two nodes. node1 and node2 mean the ID numbers of target nodes. subs means the conditions on equalities for this mutual exclusion. Namely, this mutual exclusiveness will be activated only when these equalities are satisfied.
• mutual_exclusive_edges Provides mutual exclusions between two edges. num means the number of mutual exclusive edges pairs in the latent hypotheses set.
  • xor Provides information of mutual exclusion between two edges. node1 and node2 mean the ID numbers of target edges.

Integer Linear Programming problem

Using the option -p path_ilp_out=<PATH>, you can get the detail of the ILP problem. Its elements are as follows:

• ilp The root. name generally corresponds to given observation. maximize means whether this problem maximize the objective function, or minimize it. time means the time which is spent on the conversion to ILP problem. timeout means whether the conversion have been interrupted by timeout.
• variables Provides information of all ILP variables. num means the number of ILP variables.
  • variable Provides information of an ILP variable. index corresponds the ID number of the variable. name means the variable's name. coefficient means the coefficient assigned to the variable in the objective function. If this variable is a constant, fixed attribute is added, which provides the fixed value. For example, since an observed literal must be included in the solution hypothesis, the variable corresponding to it has fixed value 1.
• constraints Provides information of all ILP constraints. num means the number of ILP constraints.
  • constraint index corresponds the ID number of the constraint. name means the constraint's name.

Solution in Integer Linear Programming

Using the option -p path_sol_out=<PATH>, you can get the detail of the solution in ILP.

Proof graph

After inference, Phillip outputs the solution hypothesis to stdout. Using the option -p path_out=<PATH>, Phillip also outputs it to give file path.