The Evolution of Neural Plasticity in Digital Organism

Learning is a phenomenon that organisms throughout nature demonstrate and that machine learning aims to replicate. In nature, it is neural plasticity that allows an organism to integrate the outcomes of their past experiences into their selection of future actions. While neurobiology has identified some of the mechanisms used in this integration, how the process works is still a relatively unclear and highly researched topic in the cognitive science field. Meanwhile in the field of machine learning, researchers aim to create algorithms that are also able to learn from past experiences; this endeavor is complicated by the lack of understanding how this process takes place within natural organisms.;In this dissertation, I extend the Markov Brain framework which consists of evolvable networks of probabilistic and deterministic logic gates to include a novel gate type-feedback gates. Feedback gates use internally generated feedback to learn how to navigate a complex task by learning in the same manner a natural organism would. The evolutionary path the Markov Brains take to develop this ability provides insight into the evolution of learning. I show that the feedback gates allow Markov Brains to evolve the ability to learn how to navigate environments by relying solely on their experiences. In fact, the probabilistic logic tables of these gates adapt to the point where the an input almost always results in a single output, to the point of almost being deterministic. Further, I show that the mechanism the gates use to adapt their probability table is robust enough to allow the agents to successfully complete the task in novel environments. This ability to generalize to the environment means that the Markov Brains with feedback gates that emerge from evolution are learning autonomously; that is without external feedback. In the context of machine learning, this allows algorithms to be trained based solely on how they interact with the environment. Once a Markov Brain can generalize, it is able adapt to changing sets of stimuli, i.e. reversal learn. Machines that are able to reversal learn are no longer limited to solving a single task. Lastly, I show that the neuro-correlate is increased through neural plasticity using Markov Brains augmented with feedback gates. The measurement of is based on Information Integration Theory and quantifies the agent's ability to integrate information.