Multidimensional Scaling by Means of Pseudoinverse Operations

Krak, Iurii; Kudin, G. I.; Kulyas, A. I.

doi:10.1007/s10559-019-00108-9

Cited by 15 publications

(5 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To evaluate the validity of our approach for the stance-detection task, we carried out computational experiments using the FNC-1 dataset and compared our findings with the leading approaches in identifying the stance of texts. This comparison included: (i) the original FNC-1 challenge baseline [42] that used feature engineering, (ii)-(iv) the top three systems from the FNC-1 challenge [43][44][45] as determined by their confusion matrices, (v) a recent study by Zhang et al [46] that had notable success with the FNC-1 dataset, (vi) current state-of-the-art by Sepulveda-Torres et al [40] in this task using the FNC-1 dataset, and (vii) our proposed approach that works by pulling out weights from the penultimate layer of the RoBERTa model and putting together the transition matrix T. The key metrics for comparison include the F 1 -score (21) for four stance categories ("Agree", "Disagree", "Discuss", and "Unrelated") and two overall performance indicators: the macro-average F 1 m (22) and weighted accuracy (23). The results shown in Table 5 are from applying all these approaches to the FNC-1 dataset.…”

Section: Interpreting Stance Detection Based On the Fnc-1 Datasetmentioning

confidence: 99%

“…Visualization aids in uncovering patterns, trends, and correlations that might otherwise go unnoticed. Notably, multidimensional scaling (MDS) [23] is considered an effective VA tool, allowing for the visualization of multidimensional data based on their similarity, presenting information about the distances between objects as points in an abstract space. Furthermore, the concept of HITL is used to denote human participation in the process of aiding the computer to make accurate decisions during model development [24].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Explainable Deep Learning: A Visual Analytics Approach with Transition Matrices

Radiuk,

Barmak,

Manziuk

et al. 2024

Mathematics

View full text Add to dashboard Cite

The non-transparency of artificial intelligence (AI) systems, particularly in deep learning (DL), poses significant challenges to their comprehensibility and trustworthiness. This study aims to enhance the explainability of DL models through visual analytics (VA) and human-in-the-loop (HITL) principles, making these systems more transparent and understandable to end users. In this work, we propose a novel approach that utilizes a transition matrix to interpret results from DL models through more comprehensible machine learning (ML) models. The methodology involves constructing a transition matrix between the feature spaces of DL and ML models as formal and mental models, respectively, improving the explainability for classification tasks. We validated our approach with computational experiments on the MNIST, FNC-1, and Iris datasets using a qualitative and quantitative comparison criterion, that is, how different the results obtained by our approach are from the ground truth of the training and testing samples. The proposed approach significantly enhanced model clarity and understanding in the MNIST dataset, with SSIM and PSNR values of 0.697 and 17.94, respectively, showcasing high-fidelity reconstructions. Moreover, achieving an F1m score of 77.76% and a weighted accuracy of 89.38%, our approach proved its effectiveness in stance detection with the FNC-1 dataset, complemented by its ability to explain key textual nuances. For the Iris dataset, the separating hyperplane constructed based on the proposed approach allowed for enhancing classification accuracy. Overall, using VA, HITL principles, and a transition matrix, our approach significantly improves the explainability of DL models without compromising their performance, marking a step forward in developing more transparent and trustworthy AI systems.

show abstract

Section: Interpreting Stance Detection Based On the Fnc-1 Datasetmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Explainable Deep Learning: A Visual Analytics Approach with Transition Matrices

Radiuk,

Barmak,

Manziuk

et al. 2024

Mathematics

View full text Add to dashboard Cite

show abstract

“…Various algorithms based on conventional computer vision with hand-crafted features, such as the histogram of oriented gradients [7], bag-of-features [8], hyperplanes separation [9]. Unfortunately, due to mentioned before, peculiarities of the hand and it's high-dynamic nature, such approaches are not robust enough and suffer a lot from changes in environment, background, or quality of the input images.…”

Section: Existing Researchmentioning

confidence: 99%

Using the Temporal Data and Three-dimensional Convolutions for Sign Language Alphabet Recognition

Kondratiuk¹,

Krak²,

Kuznetsov³

et al. 2022

Computer Modeling and Intelligent Systems

View full text Add to dashboard Cite

“…The method of MDS allows to place objects in a space of some small dimension (in this case, it is equal to two) in order to reproduce the observed distances between them with the smallest error. Thus, representing the visibility of the location of objects in the generalized two‐dimensional space of generalized features 64,65 …”

Section: Separating N‐dimensional Objects By Hyperplanes For Classifi...mentioning

confidence: 99%

“…Thus, representing the visibility of the location of objects in the generalized two-dimensional space of generalized features. 64,65 Furthermore, data visualization is used as a way of displaying a multidimensional distribution of data on a two-dimensional plane, in which the basic patterns inherent in the original distribution are qualitatively displayed. At the same time, it is necessary to minimize the loss of information content and its manifestations in the cluster structure, topological features, and dependencies between the characteristics of the location of the data in the original space.…”

Section: Introduction and Statement Of The Problem Classification Bas...mentioning

confidence: 99%

Using visual analytics to develop human and machine‐centric models: A review of approaches and proposed information technology

Krak

Barmak

Manziuk

2020

Computational Intelligence

View full text Add to dashboard Cite

The use of a visual analytical system in machine learning is the basis for the integration of human and the use of his intellectual capabilities in the construction of models. At the same time, visual analytics is used to expand human knowledge and is used as a research tool. We investigate the forms and goals of using visual analytics workflow toward the formation of the final product. Workflow is divided into human-oriented and machine-oriented in order to build a model as an information processor and decision-making mechanism. Models are built on the basis of the end user, which can be either a machine or a human. The concepts of model building and the role of machines and humans in these processes are investigated. A practical implementation of the classification information technology in the studied concept "using opposite model" in the machine-oriented visual analytics workflow for using the machine model is proposed. The basis for this model is a model formed and used by human. To classify data, human intellectual abilities are used. The boundaries of classes are determined by a human and then projected into a hyperspace of attributes with the formation of a classification model that the machine uses. Information

show abstract

Multidimensional Scaling by Means of Pseudoinverse Operations

Cited by 15 publications

References 12 publications

Explainable Deep Learning: A Visual Analytics Approach with Transition Matrices

Explainable Deep Learning: A Visual Analytics Approach with Transition Matrices

Using the Temporal Data and Three-dimensional Convolutions for Sign Language Alphabet Recognition

Using visual analytics to develop human and machine‐centric models: A review of approaches and proposed information technology

Contact Info

Product

Resources

About