Sentient Solves the 70-Multiplexer Problem

Sentient Solves the 70-Multiplexer Problem

As a follow up to their paper on tackling the Boolean Multiplexer problem using a distributed age-layered method, Sentient’s research team today announced convergence on the 70-multiplexer problem, a version of the problem with a search space so vast, it has never been solved using a direct method until now.

In recent years, there have been reports of higher order multiplexer problems having been solved using indirect methods that essentially make use of solutions from the lower orders, but this is the first instance we know of for this classic machine learning problem being solved at such a high order”, said Hormoz Shahrzad, principal researcher at Sentient.

The search space for the 70-multiplexer problem is extremely large (2^2^70), with comparatively minute local continuities. Koza, in his 1990 paper, showed how Genetic Programming can solve the 11-Multiplexer problem, which is impossible to solve using random search.

Sentient Solves the 70-Multiplexer Problem2

Here are two rule-sets, evolved by the Sentient AI platform, that are solutions to the 70-Multiplexer problem. Solving the 70-Multiplexer problem is further evidence of the power of the Sentient AI platform and validates the hub and spoke approach to distributing Evolutionary algorithms, and the superiority of the age-layered feature.

In the near future, we plan to take on the harder 135-Multiplex problem, so stay tuned!