Abstract
Consider a function where its entries are distributed among many parties. Suppose each party is allowed to transmit only a limited amount of information to a net. One can use a classical protocol to guess the value of the global function. Is there a quantum protocol improving the results of all classical protocols? Brukner et. al. showed the deep connection between such problems and the theory of Bell's inequalities. Here we generalize the theory to trits. There the best classical protocol fails whereas the quantum protocol yields the correct answer.