PU-UNet integrates stabilized product units into low-resolution residual blocks of a U-Net, reporting higher Dice scores than a matched residual U-Net baseline on ISIC 2018, Kvasir-SEG, and BUSI datasets with nearly identical parameters and latency.
arXiv preprint arXiv:2505.04397 (2025)
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
years
2026 3verdicts
UNVERDICTED 3representative citing papers
Complex AM-PU-GRU achieves 0.227 MeV interpolation and 0.179 MeV extrapolation RMSE on AME2016/AME2020 nuclear mass data, outperforming real-valued GRU and other ML models while remaining robust to WS4 and SEMF priors.
PURe networks combine product units with residuals to explicitly capture cross-feature couplings, yielding competitive accuracy plus gains in robustness and interaction coherence versus MLP baselines.
citing papers explorer
-
Product units in gated recurrent units improve nuclear-mass prediction
Complex AM-PU-GRU achieves 0.227 MeV interpolation and 0.179 MeV extrapolation RMSE on AME2016/AME2020 nuclear mass data, outperforming real-valued GRU and other ML models while remaining robust to WS4 and SEMF priors.
-
Modeling Nonlinear Feature Interactions with Product-Unit Residual Networks
PURe networks combine product units with residuals to explicitly capture cross-feature couplings, yielding competitive accuracy plus gains in robustness and interaction coherence versus MLP baselines.