MIT News May 9, 2023
Unlike humans, computer vision models don’t typically exhibit perceptual straightness, so they learn to represent visual information in a highly unpredictable way. But if machine-learning models had this ability, it might enable them to better estimate how objects or people will move. Researchers at MIT explored the relationship between network architecture, differing types of robustness, biologically-inspired filtering mechanisms, and representational straightness in response to time-varying input; they identified strengths and limitations of straightness as a useful way of evaluating neural network representations. They found that adversarial training leads to straighter representations in both convolutional neural network (CNN) and transformer-based architectures, but this effect is task-dependent, not generalizing to tasks such as segmentation and frame-prediction, where straight representations are not favorable for predictions, and to other types of robustness. The straighter representations imparted temporal stability to class predictions, even for out-of-distribution data. Biologically inspired elements increased straightness in the early stages of a network but did not guarantee increased straightness in downstream layers of CNNs. They showed that straightness is an easily computed measure of representational robustness and stability, as well as a hallmark of human representations with benefits for computer vision models… read more. Open Access TECHNICAL ARTICLEÂ