Deep Learning for Design
Deadline for submissions: 31st August 2019 (extended!)
Machine learning is generally concerned with the development of computational systems that can improve their performance of a task with experience (typically in the form of data). Deep learning is a subset of machine learning that seeks to improve performance and robustness of computational systems by imbuing them with the ability to construct internal representations of the data at varying degrees of abstraction. Deep neural networks are an increasingly common method within these broader categories. Over the last decade they have proven particularly useful for a variety of tasks including image recognition, sparking a variety of applications to related tasks in design. These have spanned architecture, mechanical design, product design, and game design.
However, the relationship that these and other applications of deep learning have to the existing human stakeholders in design needs further exploration. In what way can deep learning augment or automate the behaviors of designers? In what way can it allow designers to better capture and respond to the needs, states, or behaviors of users and customers? How does deep learning connect to existing theoretical design frameworks?
The aim of this thematic collection is to bring together leading edge research that focuses on the intersection of design and deep learning. Design Science is an archival publication venue that has a multidisciplinary readership. Since it is an online journal, papers in a Thematic Collection are published immediately upon acceptance. In keeping with open access principles, contributors are encouraged to make relevant software for their submission publicly accessible (e.g., as a GitHub repository or a Journal of Open-Source Software submission).
Automating evaluation, analysis of potential solutions
Enabling robust yet efficient simulations of complex systems
Creating rich virtual worlds
Automating the synthesis of novel solution concepts
Facilitating intelligent interactive systems via machine learning constructs
Engaging in human/machine co-creative design
Leveraging deep learning for improved agent-based systems
Generating unique design representations afforded by deep learning
For additional topics and other questions regarding this collection, please contact the Guest Editors.
Christopher McComb, Penn State University
Kazjon Grace, University of Sydney
Akin Kazakci, MINES ParisTech
C. Tucker, "Latent Space Exploration of Deep Generative Design Models"
G. Williams, C. McComb, "From Form to Function and Back Again"
A. Kazakci, M. Cherti, B. Kegl, "Digits That Are Not: Generating New Types Through Deep Neural Nets"
N. Manajan, M. Li, J. Menold, C. McComb, "Evaluating Human Trust in Deep Learning"
S. Singaravel, P. Geyer, "Interpretation of Hidden Layer Representation for Design Decisions"
W. Chen, Z. Boukouvalas, M. Fuge, "Learning Structured Design Space Representations: Unifying Generative Models of Design via Riemannian Geometry"
For all questions on the Design Science journal please contact Panos Y. Papalambros, Editor in Chief, or John S. Gero, Co-Editor in Chief.