Hyperdimensional computing for structured querying on tabular data embeddings
Researchers have explored the application of Hyperdimensional Computing (HDC), specifically the Holographic Reduced Representations (HRR) model, for structured querying on tabular data embeddings. This approach aims to address the limitation of current embedding methods where similarity scores lack intrinsic meaning, making it difficult to set reliable thresholds for retrieval and detect cases where no valid answer exists. The study demonstrates that HDC can match or outperform existing graph-based baselines in row retrieval tasks, offers more robust handling of non-equality predicates, and uniquely provides principled thresholds for identifying zero-match scenarios. AI