A previous research has shown that most learning strategies fail to
learn relational concepts when descriptions involving more then
three variables are required.
The reason resides in the emergence of the phase transition in the
covering test.
The presence of the phase transition in the covering test influences
the complexity of the search, the complexity of the hypothesis found
by the learner and the probability of success.
To avoid these limitations we have proposed an alternative learning
strategy, combining a Monte Carlo stocastich search with local
deterministic search.