My name is Marin Vogelsang and I’m a JSPS Postdoctoral Fellow working with Prof. Pawan Sinha at the Department of Brain and Cognitive Sciences at MIT since April 2024. With a background in Computational Neuroscience, Computer Science, and Cognitive Science, my work is currently focussed on studying visual learning in humans and machines. I am primarily engaging in simulations with deep neural networks as computational model systems but am also fortunate to be involved in the study of children treated for congenital blindness late in life, through Project Prakash.
Human visual recognition is remarkably robust to changes in color information. Here, we provide a potential account of the roots of this resilience based on observations with congenitally blind children who gained sight late in life. Following their sight-restoring surgeries, the removal of color cues markedly reduced their recognition performance, whereas age-matched normally sighted children showed no such decrement. This finding may be explained by the greater-than-neonatal maturity of the late-sighted children’s color system at sight onset, inducing overly strong reliance on chromatic cues. Simulations with deep neural networks corroborate this hypothesis. These findings highlight the adaptive significance of typical development and provide guidelines for enhancing machine vision systems.
Towards the end of the second trimester of gestation, a human fetus is able to register environmental sounds. This in utero auditory experience is characterized by comprising strongly low-pass-filtered versions of sounds from the external world. In this paper, we present results of computational simulations supporting the view that exposure to initially degraded auditory inputs serves an adaptive purpose – it may induce the neural development of extended temporal integration, which is crucial for tasks such as emotion recognition. In addition, this finding can help explain some of the auditory impairments associated with preterm births, suggests guidelines for the design of auditory environments in neonatal care units, and points to enhanced training procedures for computational models.
Human perceptual development evolves in a stereotyped fashion, with initially limited capabilities maturing over the months or years following the commencement of sensory experience into robust proficiencies. Here, we review recent findings from studies of children who experienced alterations of early development, as well as results from computational models, which provide compelling evidence that the limitations of early sensory experience may act as scaffolds, rather than hurdles, being causally responsible for the acquisition of later perceptual proficiencies. These results have implications for understanding typical developmental trajectories, help account for the perceptual deficits observed in atypically developed individuals, and inspire more robust training procedures for computational learning systems.