LAUSR

Invertible logic can realize bidirectional operations of a function represented by a Hamiltonian (energy) with noise and can be applied for integer factorization and training neural networks. However, the computation error is not negligible due to becoming trapped at the local minimum energy when the Hamiltonian is large. This paper introduces self-adaptive gate control (SAGC) for high convergence rates to the global minimum energy on invertible logic. From our analysis based on simulating small-scale invertible logic circuits, it is supposed that becoming stuck is caused by invalid states of invertible gates. The proposed SAGC autonomously detects invalid gates by checking the truth tables and adds large noise to them for efficient escape from the local minimum energy. As a typical example of invertible logic, invertible adders, multipliers and multiplexers are designed and evaluated. The simulation results show that the convergence probabilities to the global minimum based on SAGC are a few times higher than those based on a conventional method that equally adds noise to all the gates.