A Deep Reinforcement Learning Model-based Optimization Method for Graphic Design

Qi Guo, Zhen Wang


The significance of Deep Reinforcement learning is sensibly represented in the method of optimizing the graphic design and space framework of buildings in context with the worldwide big data environment, wherein people have increasingly stringent requirements for building layout and design and conventional layout is increasingly inadequate. This research put out a novel approach to topology optimization using deep learning in geometry. Deep neural networks characterize the density distribution in the design domain. By employing a geometry-based deep learning approach to represent the density distribution function, we can successfully avoid the checkerboard phenomena and ensure a smooth border. With a deep learning reinforcement approach, the design variables may be drastically decreased. In adjusting the designs of neural networks, we may fine-tune not only the minimal length but also the structural complexity. The effectiveness of the suggested technique is shown by several 2-Dimensional and 3-Dimensional numerical results ranging from minimal conformance to stress-constrained issues.

Full Text:



Request to process the paper in UF

DOI: https://doi.org/10.31449/inf.v48i5.5295

Creative Commons License
This work is licensed under a Creative Commons Attribution 3.0 License.