Research on deep learning architecture optimization method for intelligent scheduling of structural space.

Deep neural networks have demonstrated impressive performance across various fields such as computer vision, natural language processing, and autonomous systems. However, their static architectural configurations often result in computational inefficiencies and lack of flexibility, which limit their applicability in intelligent scheduling tasks involving complex structural spaces. To address these challenges, we propose a novel optimization framework for deep learning architectures that enables dynamic and knowledge-driven adaptation, specifically tailored for intelligent scheduling in structural environments. The proposed framework integrates a Dynamic Compositional Architecture (DCA) and a Knowledge-Embedded Adaptive Strategy (KEAS). DCA models the network as a directed acyclic graph composed of modular units, where each module is conditionally activated based on input semantics and spatial scheduling requirements. This enables real-time adjustment of computation depth, width, and routing pathways, allowing for efficient processing across varying structural configurations. KEAS enhances the model's adaptability by embedding symbolic domain knowledge and semantic constraints into the learning process, guiding the architecture to align with scheduling objectives and structural semantics. Experimental results on multiple benchmark datasets demonstrate that our approach not only achieves state-of-the-art prediction accuracy, but also significantly improves scheduling efficiency, reduces inference latency, and minimizes resource usage. This research offers a scalable and interpretable deep learning paradigm for intelligent scheduling of structural spaces in dynamic, resource-constrained environments.