Comparing Strategies for Visualizing the High-Dimensional Exploration Behavior of CPS Design Agents

Agrawal, Akash; McComb, Christopher

doi:10.1109/destion56136.2022.00017

Cited by 2 publications

(3 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To verify the extent of this difference, we calculated the Pearson correlation coefficient for over 2,100 combinations of different parameters' values for the UMAP and the weights for the dimensions. For the defined parameters (equal weights to all dimensions, number of neighbours 60 and minimum distance as 0.25) we obtained a Pearson correlation coefficient of 0.501, which is even better than the best performing algorithm for dimensionality reduction analysed by Agrawal and McComb (2022). This analysis indicates that although there is a distortion between the high-dimensional and embedded distances across the solutions, those seem to be in line with other approaches in the field and can be seen as acceptable for the current application.…”

Section: Example Application: a Game-based Design Problemmentioning

confidence: 81%

“…Accounting for different types of data generated and manipulated throughout the design process (e.g., verbalisations, sketches) adds yet another layer of complexity to the analysis (Goldschmidt, 2006). Existing computational approaches are usually restricted to numerical variables manipulated automatically via a guiding set of inequalities that define the design problem (Agrawal and McComb, 2022). Despite these challenges, many attempts have been made to create design space visualisations.…”

Section: Design Space Visualisationmentioning

confidence: 99%

“…Once the pairwise distance matrix for the solutions has been calculated, it becomes possible to create a 2D embedding of the solutions using dimensionality reduction techniques which include, among others, the Principal Component Analysis (PCA), the Uniform Manifold Approximation and Projection (UMAP), and the t-distributed Stochastic Neighbor Embedding (t-SNE) which are among the ones compared in Agrawal and McComb (2022) design space exploration with reinforcement learning agents. The idea behind these approaches is to infer how the different solutions are spread out in the higher dimensional space and then project those relationships onto a lower dimensional embedding (usually 2D), while (trying) to preserve local and global characteristics of the data.…”

Section: Creating the 2d Embedding Of The Solutionsmentioning

confidence: 99%

See 2 more Smart Citations

DS-Viz: a method for visualising design spaces

Paravizo,

Crilly

2024

Proc. Des. Soc.

View full text Add to dashboard Cite

Problems, solutions, and design itself have been framed as spaces in design research. Visualising the design space and how designers explore it, can give insight into the design process. This paper reports on a novel method for creating Design Space Visualisations (DS-Viz) that generates 2D and 3D representations of design spaces. We show how DS-Viz can be used to investigate designer behaviour, design processes and outcomes using a game-based design activity as an example. We discuss DS-Viz implications for design research highlighting potential benefits to design education and practice.

show abstract

Section: Example Application: a Game-based Design Problemmentioning

confidence: 81%

Section: Design Space Visualisationmentioning

confidence: 99%

Section: Creating the 2d Embedding Of The Solutionsmentioning

confidence: 99%

See 1 more Smart Citation

DS-Viz: a method for visualising design spaces

Paravizo,

Crilly

2024

Proc. Des. Soc.

View full text Add to dashboard Cite

show abstract

Explainable reinforcement learning (XRL): a systematic literature review and taxonomy

Bekkemoen

2023

Mach Learn

View full text Add to dashboard Cite

In recent years, reinforcement learning (RL) systems have shown impressive performance and remarkable achievements. Many achievements can be attributed to combining RL with deep learning. However, those systems lack explainability, which refers to our understanding of the system’s decision-making process. In response to this challenge, the new explainable RL (XRL) field has emerged and grown rapidly to help us understand RL systems. This systematic literature review aims to give a unified view of the field by reviewing ten existing XRL literature reviews and 189 XRL studies from the past five years. Furthermore, we seek to organize these studies into a new taxonomy, discuss each area in detail, and draw connections between methods and stakeholder questions (e.g., “how can I get the agent to do _?”). Finally, we look at the research trends in XRL, recommend XRL methods, and present some exciting research directions for future research. We hope stakeholders, such as RL researchers and practitioners, will utilize this literature review as a comprehensive resource to overview existing state-of-the-art XRL methods. Additionally, we strive to help find research gaps and quickly identify methods that answer stakeholder questions.

show abstract

Comparing Strategies for Visualizing the High-Dimensional Exploration Behavior of CPS Design Agents

Cited by 2 publications

References 13 publications

DS-Viz: a method for visualising design spaces

DS-Viz: a method for visualising design spaces

Explainable reinforcement learning (XRL): a systematic literature review and taxonomy

Contact Info

Product

Resources

About