Tsetlin machine

Uses your system voice; voice and rate in settings.

Snapshot

≈ 3 min read

Names in other languages

No names in other languages are held for this entry yet.

Article bodies are English and are not machine-translated. A name in another language is not a translation of the article.

A Tsetlin machine is an artificial intelligence algorithm based on propositional logic.

A Tsetlin machine is an artificial intelligence algorithm based on propositional logic. \{\underline{\Phi}, \underline{\alpha}, \underline{\beta}, F(\cdot,\cdot), G(\cdot)\}. A Tsetlin automaton has 2n states, here : \underline{\Phi} = \{\phi_1, \phi_2, \phi_3, \phi_4, \phi_5, \phi_6\} The FSM can be triggered by two input events \underline{\beta} = \{\beta_{\mathrm{Penalty}}, \beta_{\mathrm{Reward}}\} The rules of state migration of the FSM are stated as F(\phi_u, \beta_v) = \begin{cases} \phi_{u+1},& \text{if}~ 1 \le u \le 3 ~\text{and}~ v = \text{Penalty}\\ \phi_{u-1},& \text{if}~ 4 \le u \le 6 ~\text{and}~ v = \text{Penalty}\\ \phi_{u-1},& \text{if}~ 1 It includes two output actions \underline{\alpha} = \{\alpha_1, \alpha_2\} Which can be generated by the algorithm G(\phi_u) = \begin{cases} \alpha_1, & \text{if}~ 1 \le u \le 3\\ \alpha_2, & \text{if}~ 4 \le u \le 6. \end{cases} A basic Tsetlin machine takes a vector X=[x_1,\ldots,x_o] of Boolean features as input, to be classified into one of two classes, y=0 or y=1 . Together with their negated counterparts, \bar{x}_k = {\lnot} {x}_k = 1-x_k , the features form a literal set L = \{x_1,\ldots,x_o,\bar{x}_1,\ldots,\bar{x}_o\} . C_j (X)=\bigwedge_{ {\in} L_j} l = \prod_{ {\in} L_j} l . For example, the clause C_j(X)=x_1\land{\lnot}x_2=x_1 \bar{x}_2 consists of the literals L_j = \{x_1,\bar{x}_2\} and outputs iff x_1 = 1 and x_2 = 0 . The number of clauses employed is a user-configurable parameter . Half of the clauses are assigned positive polarity. The other half is assigned negative polarity. The clause outputs, in turn, are combined into a classification decision through summation and thresholding using the unit step function u(v) = 1 ~\text{if}~ v \ge 0 ~\text{else}~ 0 : \hat{y} = u\left(\sum_{j=1}^{n/2} C_j^+(X) - \sum_{j=1}^{n/2} C_j^-(X)\right). In other words, classification is based on a majority vote, with the positive clauses voting for y=1 and the negative for y=0 . The classifier \hat{y} = u\left(x_1 \bar{x}_2 + \bar{x}_1 x_2 - x_1 x_2 - \bar{x}_1 \bar{x}_2\right) , for instance, captures the XOR-relation. v = \sum_{j=1}^{n/2} C_j^+(X) - \sum_{j=1}^{n/2} C_j^-(X) approaches a user-set target for y=1 ( -T for y=0 ). If a clause is not reinforced, it does not give feedback to its Tsetlin automata, and these are thus left unchanged. In the extreme, when the voting sum equals or exceeds the target (the Tsetlin Machine has successfully recognized the input ), no clauses are reinforced. Accordingly, they are free to learn new patterns, naturally balancing the pattern representation resources.

Text derived from https://en.wikipedia.org/w/index.php?oldid=1350144411 under CC BY-SA 4.0; modifications have been made.

Sources cited in this revision

39 named source(s) extracted from the cited revision. A source with no author is shown by its publication; nothing is invented.