deepstruct -- linking deep learning and graph theory
This tool aids researchers in analyzing and designing neural network structures, but it is incremental as it builds on existing graph theory and deep learning concepts.
The paper introduces deepstruct, a tool that connects deep learning models and graph theory to impose graph structures on neural networks or extract them from trained models, supporting research in pruning, neural architecture search, automated network design, and structure analysis.
deepstruct connects deep learning models and graph theory such that different graph structures can be imposed on neural networks or graph structures can be extracted from trained neural network models. For this, deepstruct provides deep neural network models with different restrictions which can be created based on an initial graph. Further, tools to extract graph structures from trained models are available. This step of extracting graphs can be computationally expensive even for models of just a few dozen thousand parameters and poses a challenging problem. deepstruct supports research in pruning, neural architecture search, automated network design and structure analysis of neural networks.