MixQ-GNN picks per-component bit-widths for quantized GNNs and claims exact integer message passing via a new theorem, but the theorem is flawed as written.
Ll-gnn: Low-latency graph neural net- works on fpgas for high-energy physics,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Efficient Mixed Precision Quantization in Graph Neural Networks
MixQ-GNN picks per-component bit-widths for quantized GNNs and claims exact integer message passing via a new theorem, but the theorem is flawed as written.