We present a new image classification model that extends CNNs with biologically-inspired higher-order convolutions. Outperforms standard CNNs on benchmarks and shows unique representational properties, bridging neuroscience and deep learning.