ConCS: A Continual Classifier System for Continual Learning of Multiple Boolean Problems

Human intelligence can simultaneously process many tasks with the ability to accumulate and reuse knowledge. Recent advances in artificial intelligence, such as Transfer, Multitask Layered Learning, seek replicate these abilities. However, humans must specify task order, which is often difficult particularly uncertain domain This work introduces a Continual-learning system (ConCS), that given an open-ended set of problems once each solved its solution contribute solving further problems. The hypothesis Evolutionary Computation approach Learning Classifier Systems (LCSs) form this due niched, cooperative rules. A collaboration parallel LCSs identifies sets patterns linking features classes be reused related automatically. Results from distinct Boolean integer classification problems, varying interrelations, show by combining knowledge simple complex at increasing scales. 100% accuracy achieved for tested regardless order presentation. includes intractable previous approaches, e.g. n-bit Majority-on. major contribution human guidance now unnecessary determine learning order. Furthermore, automatically generates curricula most tasks.