Experimental studies of the early stages of object learning

Understanding how the human visual system learns to segregate and recognize objects in the environment is one of the fundamental challenges in neuroscience. The infant studies used to address this question, while necessary and valuable, are operationally difficult and limited in the complexity of their experimental designs. We have identified a unique population of children in India that allows us to adopt a complementary approach. According to the WHO, India is home to the world's largest population of blind children. Many of them can be given sight, but are deprived of treatment due to poverty and lack of medical facilities. In collaboration with a hospital in New Delhi, we have launched an outreach program to identify children in need of treatment and perform corrective surgeries free of charge. This initiative is beginning to create a remarkable population of children across a wide age-range who are just setting out on the enterprise of learning how to see. We have begun following the development of visual skills in these unique children to gain insights into fundamental questions regarding object concept learning and brain plasticity. The experimental data we have gathered so far have already begun challenging some long held conceptions regarding brain plasticity and time-lines of learning. We are finding that the human brain retains an impressive ability to acquire complex visual skills even after several years of congenital blindness and that this learning unfolds in a systematic sequence. These results, and the patterns of errors that the children make in our controlled experiments, have provided insights into the nature of information they use for learning to parse the world into meaningful objects. We also plan to complement these studies with non-invasive recordings of brain activity to assess topographical changes in cortical organization as a function of time past surgery, and their correlations with the behaviorally observed skills. We call this overall effort 'Project Prakash' after the Sanskrit word for light.

Ostrovsky, Y., Andalman, A. and Sinha, P. (under review). Vision following extended congenital blindness.

Ostrovsky, Y., Meyers, E. and Sinha, P. (In preparation). Parsing visual scenes via dynamic cues.

Sinha, P. (2003). Face classification following long-term visual deprivation. Annual meeting of the Vision Sciences Society, Sarasota, FL.

Computational model for learning to parse the world into objects

Complementing Project Prakash on the computational front, we are developing algorithms for automated object-concept discovery from natural video streams. Not only is this problem of great significance in visual neuroscience, it also occurs prominently in several other areas, such as computational genomics where the task is to extract common subsequences (‘objects’) across multiple strings (‘input images’). We are developing two kinds of algorithms for this task. The first works with static imagery. Our work here builds upon past research in statistical learning and string matching to create an efficient scheme for discovering commonalities across multiple inputs. Our computational simulations show that this approach successfully learns to recognize objects even when the input images are not spatially normalized, and the image quality is significantly degraded. However, success here is dependent upon the availability of partial supervision. In order to accurately model human object learning, we cannot be critically reliant on the availability of a ‘teacher’. With this in mind, we are also developing completely unsupervised concept discovery algorithms. This work incorporates empirical data from Project Prakash to provide guidelines for, and constraints on, the algorithms’ designs and computational complexity.

Bouvrie, J. V. and Sinha, P. (To appear). Object concept learning: Observations in congenitally blind children and a computational model. Neurocomputing.

Sinha, P., Balas, B. and Ostrovsky, Y. (2005). Project Dylan – A computational model of visual object concept learning. Annual meeting of the Vision Sciences Society, Sarasota, FL.

Sinha, P., Balas, B. J. and Ostrovsky, Y. (In preparation). Learning to parse the world into objects.