An exploration-based approach to computationally supported design-by-analogy using D3

2020 ◽  
Vol 34 (4) ◽  
pp. 444-457
Author(s):  
Hyeonik Song ◽  
Jacob Evans ◽  
Katherine Fu
Keyword(s):  
Data Visualization ◽  
Matrix Factorization ◽  
Topic Modeling ◽  
Nonnegative Matrix ◽  
Frame Rate ◽  
Retrieval Process ◽  
Clustering Quality ◽  
Database Size ◽  
Design By Analogy ◽  
Analogical Retrieval

AbstractComputational support for design-by-analogy (DbA) is a growing field, as it aids the process for designers looking to draw inspiration from external sources by harnessing the power of data mining and data visualization. This study presents a unique exploration-based approach for the analogical retrieval process using a computational tool called VISION (Visual Interaction tool for Seeking Inspiration based On Nonnegative Matrix Factorization). Leveraging the U.S. patent database as a source of inspiration, VISION enables designers to visualize a patent repository and explore for analogical inspiration in a user-driven manner. To achieve this, we perform hierarchical Nonnegative Matrix Factorization to generate a clustered structure of patent data and employ D3.js to visualize the patent structure in a node-link network, in which user interaction capabilities are enabled for data exploration. In this study, we also analyze the effect of data size (ranging from 100 to 3000 patents) on two performance aspects of VISION – the clustering quality of topic modeling results and the frame rate of interactive data visualization. The findings show that the tool exhibits more randomized and inconsistent topic modeling results when the database size is too small. But, increasing the database size lowers the frame rate to the point that it could diminish designers’ ability to retrieve and recall information. The scope of the work here is to present the creation of the DbA visualization tool called VISION and to evaluate its data scale limitations in order to provide a basis for developing a visual interaction tool for the analogical retrieval process during DbA.

Approaches for Supporting Exploration for Analogical Inspiration With Behavior, Material and Component Based Structural Representations of Patent Databases

2018 ◽  
Author(s):  
Hyeonik Song ◽  
Katherine Fu
Keyword(s):  
Matrix Factorization ◽  
Mechanical Design ◽  
Three Dimensional ◽  
Graph Structure ◽  
Patent Data ◽  
Retrieval Process ◽  
Design By Analogy ◽  
Computational Methodology ◽  
Analogical Retrieval ◽  
Non Negative Matrix Factorization

This paper presents an explorative-based computational methodology to aid the analogical retrieval process in design-by-analogy practice. The computational methodology, driven by Non-negative Matrix Factorization (NMF), iteratively builds a hierarchical repositories of design solutions within which clusters of design analogies can be explored by designers. In the work, the methodology has been applied on a large repository of mechanical design related patents, processed to contain only component-, behavior-, or material-based content, to demonstrate that unique and valuable attribute-based analogical inspiration can be discovered from different representations of patent data. For explorative purposes, the hierarchical repositories have been visualized with a three-dimensional hierarchical structure and two-dimensional bar graph structure, which can be used interchangeably for retrieving analogies. This paper demonstrates that the explorative-based computational methodology provides designers an enhanced control over design repositories, empowering them to retrieve analogical inspiration for design-by-analogy practice.

scholarly journals Local Topic Discovery via Boosted Ensemble of Nonnegative Matrix Factorization

2017 ◽  
Author(s):  
Sangho Suh ◽  
Jaegul Choo ◽  
Joonseok Lee ◽  
Chandan K. Reddy
Keyword(s):  
Matrix Factorization ◽  
Topic Modeling ◽  
Large Scale ◽  
Nonnegative Matrix Factorization ◽  
Nonnegative Matrix ◽  
Gradient Boosting ◽  
Ensemble Model ◽  
High Quality ◽  
Residual Matrix ◽  
The Given

Nonnegative matrix factorization (NMF) has been increasingly popular for topic modeling of large-scale documents. However, the resulting topics often represent only general, thus redundant information about the data rather than minor, but potentially meaningful information to users. To tackle this problem, we propose a novel ensemble model of nonnegative matrix factorization for discovering high-quality local topics. Our method leverages the idea of an ensemble model to successively perform NMF given a residual matrix obtained from previous stages and generates a sequence of topic sets. The novelty of our method lies in the fact that it utilizes the residual matrix inspired by a state-of-the-art gradient boosting model and applies a sophisticated local weighting scheme on the given matrix to enhance the locality of topics, which in turn delivers high-quality, focused topics of interest to users.

Optimizing modularity with nonnegative matrix factorization

2021 ◽  
pp. 2150142
Author(s):  
Zhenhai Chang ◽  
Zhong-Yuan Zhang ◽  
Huimin Cheng ◽  
Chao Yan ◽  
Xianjun Yin
Keyword(s):  
Community Structure ◽  
Matrix Factorization ◽  
Nonnegative Matrix Factorization ◽  
Nonnegative Matrix ◽  
Frobenius Norm ◽  
Structure Detection ◽  
Clustering Quality ◽  
Modularity Maximization ◽  
And Function

Community structure detection is one of the fundamental problems in complex network analysis towards understanding the topology structure and function of the network. Modularity is a criterion to evaluate the quality of community structures, and optimization of this quality function over the possible divisions of a network is a sensitive detection method for community structure. However, the direct application of this method is computationally costly. Nonnegative matrix factorization (NMF) is a widely used method for community detection. In this paper, we show that modularity maximization can be approximately reformulated under the framework of NMF with Frobenius norm, especially when [Formula: see text] is large. A new algorithm for detecting community structure is proposed based on the above finding. The new method is compared with four state-of-the-art methods on both synthetic and real-world networks, showing its higher clustering quality over the existing methods.

Sign in / Sign up

Export Citation Format

Share Document