FloatSOM provides a distributed, out-of-core, topology-flexible GPU implementation of self-organizing maps that achieves lower quantization error than prior baselines on large benchmarks and trains a 1024-node map on one billion samples in under seven minutes on eight GPUs.
Damminda Alahakoon, Saman Halgamuge, and Srinivasan Bala
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
FloatSOM: GPU-Accelerated, Distributed, Topology-Flexible Self-Organizing Maps
FloatSOM provides a distributed, out-of-core, topology-flexible GPU implementation of self-organizing maps that achieves lower quantization error than prior baselines on large benchmarks and trains a 1024-node map on one billion samples in under seven minutes on eight GPUs.