Graph neural networks on bipartite graphs predict active sets to warm-start the DAQP dual active-set QP solver, reducing iterations and generalizing to unseen problem sizes better than MLP baselines.
Rajiv Sambharya and Bartolomeo Stellato
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Warm-starting active-set solvers using graph neural networks
Graph neural networks on bipartite graphs predict active sets to warm-start the DAQP dual active-set QP solver, reducing iterations and generalizing to unseen problem sizes better than MLP baselines.