Skip to main content

Visual object population codes
relating human brains to nonhuman and computational models with representational similarity analysis

Final Report Summary - OBJECTPOPCODESIMMM (Visual object population codes relating human brains to nonhuman and computational models with representational similarity analysis)

The project developed and applied a novel method for understanding brain representations and how the brain processes information on the basis of high-resolution functional magnetic resonance imaging. The method of representational similarity analysis (RSA) was developed and the software for it shared freely with the research community. The method has been widely adopted by labs around the world, leading to hundreds of research papers that use the technique. A breakthrough finding of the project was that deep convolutional neural networks provide by far the best current computational model of human and nonhuman primate visual object recognition. The project helped integrate computational neuroscience with systems neuroscience, providing a novel bridge between theory and experiment.