Skull-closed Autonomous Development: WWN-7 Dealing with Scales

The Where-What Networks (WWNs) consist of a series of embodiments of a general-purpose brain-inspired network called Developmental Network (DN). WWNs model the dorsal and ventral two-way streams that converge to, and also receive information from, specific motor areas in the frontal cortex. Both visual detection and visual recognition tasks were trained concurrently by such a single, highly integrated network, through autonomous development. By “autonomous development”, we mean that not only that the internal (inside the “skull”) self-organization is fully autonomous, but the developmental program that regulates the growth and adaptation of computational network is also task non-specific. This paper focused on the “skull-closed” WWN-7 in dealing with different object scales. By “skull-closed”, we mean that the brain inside the skull, except the brain’s sensory ends and motor ends, is off limit throughout development to all teachers in the external physical environment. The concurrent presence of multiple learned concepts from many object patches is an interesting issue for such developmental networks in dealing with objects of multiple scales. Moreover, we will show how the motor initiated expectations through top-down connections as temporal context assist the perception in a continuously changing physical world, with which the network interacts. The inputs to the network are drawn from continuous video taken from natural settings where, in general, everything is moving while the network is autonomously learning.