Model metamers illuminate divergences between biological and artificial neural networks

Abstract

Deep neural network models of sensory systems are often proposed to learn representational transformations with invariances like those in the brain. To reveal these invariances we generated "model metamers" - stimuli whose activations within a model stage are matched to those of a natural stimulus. Metamers for state-of-the-art supervised and unsupervised neural network models of vision and audition were often completely unrecognizable to humans when generated from deep model stages, suggesting differences between model and human invariances. Targeted model changes improved human-recognizability of model metamers, but did not eliminate the overall human-model discrepancy. The human-recognizability of a models metamers was well predicted by their recognizability by other models, suggesting that models learn idiosyncratic invariances in addition to those required by the task. Metamer recognition dissociated from both traditional brain-based benchmarks and adversarial vulnerability, revealing a distinct failure mode of existing sensory models and providing a complementary benchmark for model assessment.