Learning to Trace Seiberg Dualities
A machine learning study uses transformers and multi-layer perceptrons to identify Seiberg dualities in supersymmetric quiver gauge theories. For quivers with around 10 nodes, these networks outperform deterministic algorithms. The work also incorporates pathfinder algorithms to enhance performance.
Researchers applied machine learning to the problem of establishing Seiberg dualities between supersymmetric quiver gauge theories, which mathematically involves quiver mutations. They found that transformer and MLP architectures can outperform deterministic algorithms for quivers with a modest number of nodes (order 10). The study also explores how different network architectures learn to trace these dualities and supplements the networks with pathfinder algorithms.
The paper demonstrates that neural networks, particularly transformers, can learn to perform quiver mutations more efficiently than traditional deterministic methods for small quivers. This suggests that learned heuristics may capture patterns in duality transformations that are not easily encoded in rule-based algorithms.
This research highlights the potential of machine learning to accelerate theoretical physics computations, which could lead to tools that assist physicists in exploring dualities and other complex mathematical structures. It may inspire similar approaches in other areas of mathematical physics.
While primarily a theoretical contribution, the development of efficient algorithms for duality detection could eventually impact fields like string theory and quantum field theory, where such dualities are fundamental. It may also lead to software tools for researchers.
Future work may extend these methods to larger quivers and more complex dualities, potentially integrating with symbolic computation systems. The approach could also be applied to other problems involving graph transformations or knot theory.