MO-PaDGAN: Reparameterizing Engineering Designs for augmented multi-objective optimization

Chen, Wei; Ahmed, Faez

doi:10.1016/j.asoc.2021.107909

Cited by 28 publications

(17 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Many different approaches have attempted to improve the performance of low-fidelity surrogate models. For example, self-supervised data augmentation [79] uses the DGM itself to generate samples in sparse regions (through optimization [166] or conditioning [79]) which then can be used to train the low-fidelity models to perform more accurately and in turn improve methods that rely on them for design generation. Multi-fidelity modeling is another promising approach, which involves generating surrogate models that augment a few costly high-fidelity samples with low-fidelity samples to attain higher fidelity surrogates with minimal expense.…”

Section: Design Performance Evaluationmentioning

confidence: 99%

“…Several recent works [166,19,160] have proposed methods to encourage creativity and novelty in DGMs. In their framework, CreativeGAN, Nobari et al, focus on identifying novelty and guiding DGMs towards such behaviour by directly introducing novel features into typical designs, thereby expanding the design space and novelty of the DGM's data.…”

Section: Creativity and Noveltymentioning

confidence: 99%

See 1 more Smart Citation

Deep Generative Models in Engineering Design: A Review

Regenwetter¹,

Nobari²,

Ahmed³

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Automated design synthesis has the potential to revolutionize the modern human design process and improve access to highly optimized and customized products across countless industries. Successfully adapting generative Machine Learning to design engineering may be the key to such automated design synthesis and is a research subject of great importance. We present a review and analysis of Deep Generative Learning models in engineering design. Deep Generative Models (DGMs) typically leverage deep networks to learn from an input dataset and learn to synthesize new designs. Recently, DGMs such as Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), feedforward Neural Networks (NNs) and certain Deep Reinforcement Learning (DRL) frameworks have shown promising results in design applications like structural optimization, materials design, and shape synthesis. The prevalence of DGMs in Engineering Design has skyrocketed since 2016. Anticipating continued growth, we conduct a review of recent advances with the hope of benefitting researchers interested in DGMs for design. We structure our review as an exposition of the algorithms, datasets, representation methods, and applications commonly used in the current literature. In particular, we discuss key works that have introduced new techniques and methods in DGMs, successfully applied DGMs to a design-related domain, or directly supported development of DGMs through datasets or auxiliary methods. We further identify key challenges and limitations currently seen in DGMs across design fields, such as design creativity, handling complex constraints and objectives, and modeling both form and functional performance simultaneously. In our discussion we identify possible solution pathways as key areas on which to target future work.

show abstract

Section: Design Performance Evaluationmentioning

confidence: 99%

Section: Creativity and Noveltymentioning

confidence: 99%

Deep Generative Models in Engineering Design: A Review

Regenwetter¹,

Nobari²,

Ahmed³

2021

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…They, therefore take the approach of maximizing entropy in the generated samples. Others have shown that simply promoting diversity in GANs will allow GANs to fill in data distribution gaps and even deviate slightly from the original distribution [10,9]. In this work we take a different approach and promote self creativity in GANs.…”

Section: The Creativity Problem In Gansmentioning

confidence: 99%

“…To overcome this problem, data-driven methods such as generative adversarial networks (GANs) [7] and variational autoencoders (VAEs) [8], have been employed in many design synthesis problems [9,10,11,12,13,14,15]. GANs and VAEs are generally capable of learning complex distributions of existing designs and even considering performance and quality evaluation when generating new designs [9,10]. They allow for learning an underlying low-dimensional latent space that can represent the existing designs.…”

Section: Introductionmentioning

confidence: 99%

“…Although these methods help with mode collapse, the nature of GANs methods will always tend towards data used to train them. This "emulative" [18] property of GANs is understood well, and it has been shown that GANs will rarely produce designs outside the original distribution of data [10,9,18] In addition to those rare designs, they can also produce designs that interpolate existing data and fill large gaps in the design space not filled by existing designs [10,9]. The rarity of such GAN-generated samples means they are not very useful and will likely be ignored.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

CreativeGAN: Editing Generative Adversarial Networks for Creative Design Synthesis

Nobari¹,

Rashad²,

Ahmed³

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Modern machine learning techniques, such as deep neural networks, are transforming many disciplines ranging from image recognition to language understanding, by uncovering patterns in big data and making accurate predictions. They have also shown promising results for synthesizing new designs, which is crucial for creating products and enabling innovation. Generative models, including generative adversarial networks (GANs), have proven to be effective for design synthesis with applications ranging from product design to metamaterial design. These automated computational design methods can support human designers, who typically create designs by a time-consuming process of iteratively exploring ideas using experience and heuristics. However, there are still challenges remaining in automatically synthesizing 'creative' designs. GAN models, however, are not capable of generating unique designs, a key to innovation and a major gap in AI-based design automation applications. This paper proposes an automated method, named CreativeGAN, for generating novel designs. It does so by identifying components that make a design unique and modifying a GAN model such that it becomes more likely to generate designs with identified unique components. The method combines state-of-art novelty detection, segmentation, novelty localization, rewriting, and generative models for creative design synthesis. Using a dataset of bicycle designs, we demonstrate that the method can create new bicycle designs with unique frames and handles, and generalize rare novelties * Address all correspondence to this author. to a broad set of designs. Our automated method requires no human intervention and demonstrates a way to rethink creative design synthesis and exploration.

show abstract