SARN: Relational Reasoning through Sequential Attention
This addresses a bottleneck in relational reasoning for AI systems, offering an incremental improvement over existing methods.
The paper tackled the computational inefficiency of relational networks by proposing SARN, which reduces computational and memory requirements while achieving high accuracy on the Sort-of-CLEVR dataset, particularly for relational questions.
This paper proposes an attention module augmented relational network called SARN(Sequential Attention Relational Network) that can carry out relational reasoning by extracting reference objects and making efficient pairing between objects. SARN greatly reduces the computational and memory requirements of the relational network, which computes all object pairs. It also shows high accuracy on the Sort-of-CLEVR dataset compared to other models, especially on relational questions.