Research summary
When you process visual information, your eyes send signals to your brain, which interprets them to help you decipher what you are seeing. Computer vision—a field of artificial intelligence that trains computers to interpret and understand the visual world—has been revolutionized by convolutional neural network (CNN) models that were inspired by this process. But current computational models of the visual system cannot predict neural activity in variable lighting conditions or in individuals they did not encounter in the dataset used to train them.
As Canada Research Chair in Computational Neuroscience, Dr. Joel Zylberberg aims to overcome these limitations by applying new methods from artificial intelligence known as foundation model approaches. He and his research team are also incorporating more bio-realistic components into deep learning models that can adapt to changing lighting conditions. Ultimately, Zylberberg and his team hope to develop better system models with applications in both neuroscience and computer vision.