• Publications
  • Influence
Aerodynamic Design Optimization and Shape Exploration using Generative Adversarial Networks
TLDR
This work proposes to use a deep generative model of aerodynamic designs (specifically airfoils) that reduces the dimensionality of the optimization problem by learning from shape variations in the UIUC airfoil database and shows that this model empirically accelerates optimization convergence by over an order of magnitude.
Active expansion sampling for learning feasible domains in an unbounded input space
  • Wei Chen, M. Fuge
  • Computer Science
    ArXiv
  • 25 August 2017
TLDR
This work introduces Active Expansion Sampling (AES), a method that identifies (possibly disconnected) feasible domains over an unbounded input space and shows that AES has a misclassification loss guarantee within the explored region, independent of the number of iterations or labeled samples.
BézierGAN: Automatic Generation of Smooth Curves from Interpretable Low-Dimensional Parameters
  • Wei Chen, M. Fuge
  • Computer Science
    ArXiv
  • 27 August 2018
TLDR
Results show that the proposed deep learning based generative model can generate diverse and realistic curves, while preserving consistent shape variation in the latent space, which is favorable for latent space design optimization or design space exploration.
Diverse Weighted Bipartite b-Matching
TLDR
A quadratic programming-based approach to solving a submodular minimization problem that balances diversity and total weight of the solution and defines the price of diversity, a measure of the efficiency loss due to enforcing diversity, and gives a worst-case theoretical bound.
Deep learning for molecular generation and optimization - a review of the state of the art
TLDR
A recent groundswell of work which uses deep learning techniques to generate and optimize molecules and how these techniques improve the quality of existing molecules is reviewed.
Deep learning for molecular design—a review of the state of the art
TLDR
Several important high level themes emerge, including the shift away from the SMILES string representation of molecules towards more sophisticated representations such as graph grammars and 3D representations, the importance of reward function design, the need for better standards for benchmarking and testing, and the benefits of adversarial training and reinforcement learning over maximum likelihood based training.
Airfoil Design Parameterization and Optimization using Bézier Generative Adversarial Networks
Global optimization of aerodynamic shapes usually requires a large number of expensive computational fluid dynamics simulations because of the high dimensionality of the design space. One approach ...
Synthesizing Designs With Interpart Dependencies Using Hierarchical Generative Adversarial Networks
  • Wei Chen, M. Fuge
  • Computer Science
    Journal of Mechanical Design
  • 1 November 2019
TLDR
This paper decomposes the problem of synthesizing the whole design into synthesizing each part separately but keeping the interpart dependencies satisfied, and constructs multiple generative models, the interaction of which is based on the part dependency graph.
Analysis of Collaborative Design Networks: A Case Study of OpenIDEO
TLDR
It is found that OpenIDEO’s social interactions center around a core of users who communicate more frequently with members on the periphery than among themselves (an uncommon disassortative core-periphery social structure) which is more robust to network changes than standard social networks.
...
1
2
3
4
5
...