Cognition and Neurocomputation

Abstract

Cognition and Neurocomputation is a subject that can be thought of either as a narrow specialty area of neuroscience, psychology and psychiatry or egotistically even the main area of which those other subjects are sub-areas. The underlying principle of the subject is that the emergence of the phenomenon of human cognition is approachable and even explainable as the natural emergence of a macro-phenomenon from an extensive micro-phenomenon in analogy as, for example, classical Newtonian physics can explain the motions of planets. Historically when there is a breakthrough in neurocomputation there are attempts to explain cognitive effects by analogy. Such was the case 70 years ago with the McCullough-Pitts neuron and simple neural networks, 60 years ago with the Perceptron, 40 years ago with neural networks and backpropagation, and now during the past 10 years with deep neural networks. Usually these "explanations" are somewhat naïve from a cognitive viewpoint and can occasionally be even a bit embarrassing. It is fair to say that the explanations of cognitive behavior directly from the neurocomputation model has not yet been an overwhelming success. Nonetheless, great advances have occurred; an example is the work of McClelland et al. [14]. Here the lesson is that when one can see the produced model and the human model as two instances of a common mechanism, where the interpretation of the mechanism can be completely different in each. The simpler the artificial model is then results in the largest strengthening of belief in the correctness of the mechanism in the human. For good examples, see McClelland et al. [14] and Peleg, Hazan et al. [13]. As a subject, we see that Neurocomputation and Cognition has at least three main components, "Cognition", "Neuro" and "Computation". "Cognition" can be thought of as (i) the goal, that is, neurocomputation is there to explain by computational means human cognitive abilities, or alternatively (ii) as the source, being taken as a well-known strong