Functional learning: a new approach to learning in digital circuits

Recently, learning-based techniques, which are extremely effective in finding test vectors for hard to detect faults and in detecting redundancies, have been proposed as an efficient alternative to the traditional branch-and-bound techniques for test generation. This paper presents functional learning, a new OBDD-based learning technique, that uses implication procedures. Functional learning is complete given enough time; it will determine all the uniquely implied values in the circuit from the current situation of value assignments. The most attractive feature of functional learning is its ability to extract, maintain and manipulate novel information regarding the circuit in compact OBDD representations for Boolean functions