AI

MinkowskiPE: Minkowski Positional Encoding for Spatiotemporal Perception

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 05/10/2026, 17:18:24 UTC.

Briefings

MinkowskiPE introduces a new positional encoding method combining joint temporal and spatial coordinates to parameterize Lorentz transformations, making attention scores invariant to global coordinate shifts. Evaluated on molecular dynamics and video prediction tasks, it achieved the best results on nine multi-trajectory molecular evaluations and reduced KTH video-prediction MSE by 9.9% relative to the best baseline while using approximately one-tenth as many parameters as existing approaches.

Key points

• MinkowskiPE uses joint temporal and spatial coordinates to parameterize Lorentz transformations applied to query and key features.
• Attention scores depend on position only through relative spacetime displacement, making them invariant to global translation of coordinates.

Context and limitations

Evaluation is limited to nine multi-trajectory molecular evaluations and KTH video-prediction tasks; results are not yet generalized to other spatiotemporal domains.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 09/10/2026, 17:00:15 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeib6p3nrelalloyvqoq7as5ctxs4mr5y7s6vdanpd36htgihuo5zhy