Encoding Robust Topological Signatures for Hyperdimensional Computing

In three linesHyperdimensional computing method using topological signatures (holes, RTS-invariant Zernike moments) to improve robustness against rotations, noise, and occlusions. Experiments on MNIST/EMNIST show substantial robustness gains over naive HD baseline and competitive accuracy with compact CNNs.

Summary generated by Claude — human-verified