| Abstract: | Two approaches on estimating the number of threshold functions which were recently developed by the author are discussed. Let P(K,n) denote the number of threshold functions in K-valued logic. The first approach establishes that $$P(K,n + 1) \geqslant \frac{1}{2}\left( {\mathop {K^{n - 1} }\limits_{\left\lfloor {n - 4 - 2\frac{n}{{\log _K n}}} \right\rfloor } } \right)P\left( {K,\left\lfloor {{\text{2}}\frac{n}{{\log _K n}} + 3} \right\rfloor } \right).$$ The key argument of investigation is the generalization of the result of Odlyzko on subspaces spanned by random selections of ±1-vectors. Let $E_K = \{ 0,1 \ldots ,K - 1\} $ and let E denote the set of all vectors $w_i ,i = 1, \ldots ,K^n $ , which have the form $(1,a_1 , \ldots ,a_n ),a_i \in E_K $ . Denote by $\Lambda _n (K)$ the number of all collections of different vectors $(w_{i_1 } , \ldots ,w_{i_n } ),2 \leqslant i_1 , \ldots ,i_n \leqslant \mathbb{K}^n $ , such that, for any k, $1 \leqslant k \leqslant n$ , the vector $w_{i_k } $ is minimal among all vectors from the set $E \cap {\text{span}}(w_{i_k } , \ldots ,w_{i_n } )$ . The second approach is based on topology-combinatorical techniques and allows to establish the following inequality $P(K,n) \geqslant 2\Lambda _n (K)$ . |
