Resilient k-d trees: k-means in space revisited

摘要: We propose a k-d tree variant that is resilient to pre-described number of memory corruptions while still using only linear space. While the data structure independent interest, we demonstrate its use in context high-radiation environments. Our experimental evaluation demonstrates resulting approach leads significantly higher resiliency rate compared previous results. This especially case for large-scale multi-spectral satellite data, which renders proposed well-suited operate aboard today’s satellites.