Abstract
There is ample neurobiological evidence that context-sensitive neocorticalneurons use their apical inputs as context to amplify the transmission ofcoherent feedforward (FF) inputs. However, it has not been demonstrated untilnow how this known mechanism can provide useful neural computation. Here weshow for the first time that the processing and learning capabilities of thisform of neural information processing are well-matched to the abilities ofmammalian neocortex. Specifically, we show that a network composed of suchlocal processors restricts the transmission of conflicting information tohigher levels and greatly reduces the amount of activity required to processlarge amounts of heterogeneous real-world data e.g., when processingaudiovisual speech, these local processors use seen lip movements toselectively amplify FF transmission of the auditory information that thosemovements generate and vice versa. As this mechanism is shown to be far moreeffective and efficient than the best available forms of deep neural nets, itoffers a step-change in understanding the brain's mysterious energy-savingmechanism and inspires advances in designing enhanced forms of biologicallyplausible machine learning algorithms.