Tsetlin Machine (TM) is a rule-based machine-learning algorithm comprising collectives of two-action Tsetlin Automata (TAs) that cooperatively form conjunctive logical clauses from Boolean inputs through stochastic feedback. The standard TM does not constrain clause composition, and therefore mutual exclusivity between a literal and its complement within the same clause is not inherently guaranteed, potentially leading to inflated clause counts and longer training cycles. This paper introduces TRIPOD, a three-action learning automaton that extends the TM paradigm to reduce hardware complexity, enhance energy efficiency, and improve learning reliability and resource utilisation. Unlike the standard TM model, where each TA is associated with a single literal, each TRIPOD is associated with a pair of literals, a variable and its complement, ensuring mutual exclusivity within a clause. Compared with TRIM, an earlier proposal for a three-action automaton, TRIPOD achieves comparable accuracy using 2.5–4x fewer clauses and drastically reduced training cycles.
Rudin et al. (Wed,) studied this question.