Protein Fold Recognition Using Markov Logic Networks - Mathematical Approaches to Polymer Sequence Analysis and Related Problems - page 74

1.8 ∀ x HeavilyDrinks ( x ) ⇒ CarAccidents ( x )
0.7

∀ x , y Friends ( x , y ) ⇒ ( HeavilyDrinks ( x ) ⇔ HeavilyDrinks ( x ))

Constants: Paolo (P) and Cesare (C)

Friends (P,C)

Friends (P,P) Friends (C,C)

HeavilyDrinks (P) HeavilyDrinks (C)

Friends (C,P)

CarAccidents (P) CarAccidents (C)

Fig. 4.4 Complete construction of the nodes of the ground Markov network

1.8

∀ x HeavilyDrinks ( x ) ⇒ CarAccidents ( x )

0.7 ∀ x , y Friends ( x , y ) ⇒ ( HeavilyDrinks ( x ) ⇔ HeavilyDrinks ( x ))

Constants: Paolo (P) and Cesare (C)

Friends (P,C)

Friends (P,P) Friends (C,C)

HeavilyDrinks (P) HeavilyDrinks (C)

Friends (C,P)

CarAccidents (P) CarAccidents (C)

Fig. 4.5 Connecting nodes whose predicates appear in some ground formula

1.8 ∀ x HeavilyDrinks ( x ) ⇒ CarAccidents ( x )
0.7 ∀ x , y Friends ( x , y ) ⇒ ( HeavilyDrinks ( x ) ⇔ HeavilyDrinks ( x ))

Constants: Paolo (P) and Cesare (C)

Friends (P,C)

Friends (P,P) Friends (C,C)

HeavilyDrinks (P) HeavilyDrinks (C)

Friends (C,P)

CarAccidents (P) CarAccidents (C)

Fig. 4.6 Connecting nodes whose predicates appear in some ground formula

4.3 Learning Approaches for MLNs

The first attempt to learn MLNs structure was that in [ 23 ], where the authors
used an inductive logic programming (ILP) system to learn the clauses and then
learned the weights by maximizing pseudo-likelihood [ 1 ]. In [ 17 ] another method