CliffordNet explores Clifford algebra in neural networks
The architecture combines inner-product and geometric interactions without attention or mixer layers; the report gives 77.82% accuracy on CIFAR-100 with 1.4 million parameters.
CliffordNet is presented as a neural-network architecture based on the geometric product of Clifford algebra. According to the report, it combines inner-product and geometric interactions without attention or mixer layers. The reported result is 77.82% accuracy on CIFAR-100 with 1.4 million parameters.
The proposal gives engineers an alternative to evaluate against attention- and mixer-based architectures; the cited result alone does not establish superiority on other tasks. If you use AI to study or apply the material, avoid sending internal or identifiable data: use public or previously anonymized content and check your organization’s policy.