n-bit Counters

We compare the Boolean learning algorithm implemented in BULL with five automata learning algorithms implemented in libalf. A classical example, n-bit counter, is used as the target for learning. To be more specific, we use Boolean learning algorithms to target the circuit of a counter and automata learning algorithms to target the state machine obtained by unfolding the circuit. Below we provide the log files of our experiments as a reference.

Log Files of n-bit Counters
Log Files of n-bit Counters with Reset

Random 3SAT formulae

Here we compare the performance of the Boolean learning algorithms using random 3SAT formulae of n variables. In those formulae, the ratio of the number of variables to the number of clauses is 1/4. Notice that this ratio is very close to satisfiability threshold of 3SAT formulae. Hence the chance of getting a satisfiable formula is 50%. We use a timeout period of 10 minutes. Below we provide interested users a summary file in excel, the 3SAT benchmarks, and the log files as a reference.

Summary
Random 3SAT formulae benchmarks
Log Files

experiments.txt · Last modified: 2012/01/21 03:04 by gulu