Constructing Visual Function Through Prenatal and Postnatal Learning (2005)
Although environment-driven learning can explain much of postnatal neural development, substantial organization and functional ability is present even at birth. Recent experimental discoveries of widespread spontaneous neural activity suggest that prenatal development may utilize very similar mechanisms and principles as postnatal learning, driven by internally generated sources instead of the environment. This chapter shows how this idea can explain features of the organization and function of the primary visual cortex (V1) and higher level face-processing areas. Specifically, we simulate how neural preferences for contour orientation and human faces can develop prenatally from internally generated activity and postnatally from natural image stimuli. These simulations are based on HLISSOM, a hierarchical self-organizing model of the development of topographic neural maps. The results match experimental neuroimaging and psychophysical data from newborn and older animals and humans, and provide concrete predictions about infant behavior and neural activity for future experiments. They also suggest that combining internally generated activity with a learning algorithm is an efficient way to develop complex neural machinery.
In Denis Mareschal and Mark H. Johnson and Sylvain Sirois and Michael Spratling and Michael S. C. Thomas and Gert Westermann, editors, Neuroconstructivism, Vol. 2: Perspectives and Prospects, 13-37, Oxford, UK, 2005. Oxford University Press.

James A. Bednar Postdoctoral Alumni jbednar [at] inf ed ac uk
Risto Miikkulainen Faculty risto [at] cs utexas edu