An Exploration of Explainable Machine Learning Using Semantic Web Technology

Procko, Tyler; Elvira, Timothy; Ochoa, Omar; Rio, Nicholas Del

doi:10.1109/icsc52841.2022.00029

Cited by 6 publications

(5 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…11 neurons -neuron number4,9,11,12,15,16,23,27, 60, 62, and 63 got activated by more than 90% activations in all the three cases.-10 neurons -neuron numbers 6, 13, 29, 36, 37, 39, 45, 52, 54, 59 were below 1% activations in all three cases. the rest activations are in the range of 1 -56.52%.…”

mentioning

confidence: 87%

“…Nonetheless, these approaches need to improve in terms of producing deeper explanations generated over more expressive background knowledge. [23] follows the effort of [30], by semi-automating the DL Learner tool, which provides explanations to ML algorithms using semantic background knowledge. However, while DL-Learner is a very useful system in producing theoretically correct results has significant performance issues in some scenarios, such as a single run of DL-Learner can easily take over two hours; in contrast the scenario easily necessitates thousands of such runs.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Explaining Deep Learning Hidden Neuron Activations using Concept Induction

Abhilekha¹,

Barua²,

Hitzler³

2023

Preprint

View full text Add to dashboard Cite

One of the current key challenges in Explainable AI is in correctly interpreting activations of hidden neurons. It seems evident that accurate interpretations thereof would provide insights into the question what a deep learning system has internally detected as relevant on the input, thus lifting some of the black box character of deep learning systems.The state of the art on this front indicates that hidden node activations appear to be interpretable in a way that makes sense to humans, at least in some cases. Yet, systematic automated methods that would be able to first hypothesize an interpretation of hidden neuron activations, and then verify it, are mostly missing. In this paper, we provide such a method and demonstrate that it provides meaningful interpretations. It is based on using large-scale background knowledge -a class hierarchy of approx. 2 million classes curated from the Wikipedia Concept Hierarchy -together with a symbolic reasoning approach called concept induction based on description logics that was originally developed for applications in the Semantic Web field. Our results show that we can automatically attach meaningful labels from the background knowledge to individual neurons in the dense layer of a Convolutional Neural Network through a hypothesis and verification process.

show abstract

mentioning

confidence: 87%

Section: Related Workmentioning

confidence: 99%

Explaining Deep Learning Hidden Neuron Activations using Concept Induction

Abhilekha¹,

Barua²,

Hitzler³

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…As neural networks do not offer any direct interpretation of feature importance, unlike the coefficients available from a logistic regression model or the built-in feature importance for tree-based models like random forests, we employed the SHAP (SHapley additive exPlanations) method [10] to explain how each feature affects the model. Finally, as these methods are mostly focused on the importance of features, we also employed the DL-Learner framework [11] to generate explanations, in the form of description logics sentences, which can be evaluated regarding their readability and accuracy.…”

Section: Setup and Experimentsmentioning

confidence: 99%

“…1) using the Protégé editor and imported the dataset content as instances of this ontology. After that, we configured the DL-Learner framework to employ this ontology as background knowledge, together with the Class Expression Learning for Ontology Engineering (CELOE) algorithm [22], to generate explanations. CELOE builds a search tree based on refinement operators, which define a mapping from a given input concept description to a set of derived or refined concept descriptions.…”

Section: Explanations With Description Logicsmentioning

confidence: 99%

Toward a Holistic Computational Representation for Sleep Quality and its Support for Explainability

Siebra,

Amorim,

Quintino

et al. 2023

2023 45th Annual International Conference of the IEEE Engineering in Medicine &Amp; Biology Society (EMBC)

View full text Add to dashboard Cite

Sleep quality is recognized as one of the main factors that affect human health. Thus, several studies have been encouraged to analyze features, such as stress level and female menopause, which are directly related to sleep quality. While these works rely mostly on reductionism as the philosophical framework, we approach this problem from a holist perspective, using a model with 10 features that could provide more reliable explanations for inductive conclusions. We demonstrate the principles of this hypothesis by analyzing the data regarding the day before a sleep episode of 1736 volunteers. This analysis shows, for example, the performance of each feature when they are jointly used along prediction tasks. Moreover, we evaluate the readability and accuracy of explanations, given as description logic sentences and based on a knowledge representation that considers the 10 features as elements that compose a sleep quality ontological definition.Clinical Relevance-Patient symptoms, such as sleep issues, and the consequence of their constructors must be explained as evidence in diagnosis and clinical trials. This mainly helps the resolution of potential disagreements between systems and human experts.

show abstract

“…The development and implementation of web-based elearning are increasingly varied, such as WebQuest (WQ) for question-based web learning [40], content management systems (Joomla and Moodle), website e-learning [41], [42], semantic website technology [43], WebMO for natural science, and chemistry learning [44], web and video [45], Semantic Web technology, version 3.0 [46], web-based application of synchronous and asynchronous learning [47], clinical supervision [48], webinar, and web conference learning in elementary schools [49], [50]. Several countries have incorporated web-based e-learning into their curriculum, Learning Management systems (LMS) [51], [52], models, and media for remote-controlled learning systems, online discussion rooms, and student assessment control tools [53]- [56].…”

Section: Introductionmentioning

confidence: 99%

Web-based E-learning in Elementary School: A Systematic Literature Review

Purnama,

Wilujeng,

Abdul Jabar

2023

JOIV : Int. J. Inform. Visualization

View full text Add to dashboard Cite

This article presents literature review on web-based e-learning in elementary school in the latest literature. SLR method and PRISMA protocol with the stages of identification, screening, eligibility, inclusion, and abstraction, data analysis assisted by the Publish or Perish 7 application, VOSviewer, and NVIVO 12 Plus. The results of searching for articles on Scopus through the Publish or Perish 7 application are 507. Then the articles were filtered according to compatible themes into 50 articles. The topic findings are web-based e-learning, elementary school, the impact of web-based e-learning and web-based e-learning concept, academic performance, teaching/learning strategies, online learning, Covid-19, HPC database, web-based applications, distance learning, 3D visualization, automation, strategic learning, semantic web, technology, education, linguistic content, big data architecture, learning setting, e-readiness, linguistic content, STEM, etc., that are directly or indirectly connected. The 50 articles were analyzed according to the specified topics through the NVIVO 12 Plus application, and the results were described according to the research questions. The findings in this article explain that web-based e-learning integrates pedagogy and technology and becomes part of digital multimedia implemented in e-learning, blended learning, and face-to-face that impacts elementary school students and teachers directly or indirectly. Future research needs to explore web-based e-learning in schools that is current, safe, and needed by students and teachers.

show abstract

An Exploration of Explainable Machine Learning Using Semantic Web Technology

Cited by 6 publications

References 0 publications

Explaining Deep Learning Hidden Neuron Activations using Concept Induction

Explaining Deep Learning Hidden Neuron Activations using Concept Induction

Toward a Holistic Computational Representation for Sleep Quality and its Support for Explainability

Web-based E-learning in Elementary School: A Systematic Literature Review

Contact Info

Product

Resources

About