A Lower Bound on Branching Programs Reading Some Bits Twice

By (1, + k(n))-branching programs (b.p.'s) we mean those b.p.'s which during each of their computations are allowed to test at most k(n) input bits repeatedly. For a Boolean function computable within polynomial time a trade-off is presented between the size and the number of repeatedly tested input bits of any b.p. P computing the function. Namely, if at most k(n) repeated tests are allowed, where log2 n ⩽ k(n) ⩽ n(1000 log2 n), then the size of P is at least exp(Ω(n(k(n)log2 n))12). This is exponential whenever k(n) ⩽ nα for a fixed α < 1 and superpolynomial whenever k(n) = o(nlog32 n). The presented result is a step towards a superpolynomial lower bound for 2-b.p.'s which is an open problem since 1984 when the first superpolynomial lower bounds for 1-b.p.'s were proven (Wegener, 1988; Žak, 1984). The present result is an improvement on (Žak, 1995).