Landscape of Requirements Engineering for Machine Learning-based AI Systems

Yoshioka, Nobukazu; Husen, Jati Hiliamsyah; Tun, Hnin Thandar; Chen, Zhenxiang; Washizaki, Hironori; Fukazawa, Yoshiaki

doi:10.1109/apsecw53869.2021.00011

Cited by 4 publications

(6 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Strengths, weaknesses, and challenges facing these applications were outlined. In Artificial Intelligence (AI) based systems, the current literature has focused on using AI to manage RE activities with limited research on RE for AI [20,21,22]. This domain includes mainly AI and machine learning components.…”

Section: Related Workmentioning

confidence: 99%

“…Requirements Engineering (RE) is being increasingly applied in different business domains, and its application benefits are evaluated as valuable [2,3,4,[17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32]. Despite this active RE application and its added value in different domains, the TLC domain has not yet been approached, regardless of its industrial interest.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Untitled

2024

IJNSA

View full text Add to dashboard Cite

Requirements Engineering (RE) is the most important activity and pivot phase in the software development life cycle. It consists mainly of requirements analysis, design, and specification. The application of RE in a business domain aims to generalize its different specific approaches by meta-models under which each specific approach is an instance or specialization. RE is being applied more and more in different business domains, and its application benefits are found to be valuable. Despite these active RE applications and their added values in different domains, the Traffic Light Control (TLC) domain has not yet been approached, regardless of its everyday interest. The work presented in this paper is part of a TLC enhancement project that explores RE's potential contributions to leveraging TLC quality. Thus, in this work, different recent specific approaches to TLC are analyzed, and a large part of this business domain's functional/ non-functional requirements are elicited and specified. Moreover, the traffic context-aware traffic light control systems (CTLC) are also considered. They are deeply analyzed and compared to the typical TLC systems. As a primary result, the tangling of different concerns was stated, and a separate concerns paradigm was applied. This leads to requirements specification through separated agents. This should lead to TLC and CTLC systems development and maintenance cost reduction. An important agent dealing with security aspects and their evolving technologies is introduced. The impact evaluations of the RE on some TLC and CTLC current approaches are also presented.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Untitled

2024

IJNSA

View full text Add to dashboard Cite

show abstract

“…Apesar da existência de vários trabalhos na literatura que abordam a confianc ¸a em IA, há uma escassez de pesquisas sobre requisitos para lidar com a confianc ¸a em IA/AM. Essa lacuna pode ser atribuída à falta de pesquisas sobre Engenharia de Requisitos (ER) para AM [Yoshioka et al 2021]. Abordar requisitos confiáveis é uma tarefa de grande importância para o desenvolvimento de sistemas de IA seguros e confiáveis [Kaur et al 2022].…”

Section: Introduc ¸ãOunclassified

“…Para manter e desenvolver requisitos viáveis e valiosos para sistemas de AM, é necessário existir uma colaborac ¸ão entre cientistas de dados, engenheiros de software e especialistas de domínio [Yoshioka et al 2021]. Buscar técnicas que permitam o envolvimento colaborativo na elicitac ¸ão de requisitos de sistemas de AM é crucial para o desenvolvimento e sucesso dessas aplicac ¸ões.…”

Section: Introduc ¸ãOunclassified

Brainwriting na elicitação de requisitos para IA confiável

Silva

Sampaio

Bezerra

et al. 2023

Anais Do VIII Workshop Sobre Aspectos Sociais, Humanos E Econômicos De Software (WASHES 2023)

View full text Add to dashboard Cite

Falhas apontadas em sistemas de Aprendizado de Máquina (AM) têm afetado a confiança dos usuários, tornando-se essencial abordar o tema. Este estudo investiga o uso da técnica Brainwriting na elicitação de requisitos para Inteligência Artificial (IA) confiável, com foco em AM, por meio da colaboração multidisciplinar dos papéis envolvidos. O estudo foi conduzido com um grupo de mulheres com perfis de atuação diferentes, facilitado por um mediador. Ao final do estudo, um conjunto de perguntas abertas foi aplicado para compreender a perspectiva das participantes em relação à IA confiável. Com o estudo foi possível criar requisitos para a IA confiável, com ênfase em AM, além de proporcionar a participação multidisciplinar, utilizando Brainwriting.

show abstract

“…In practice, the model pruning is mostly performed by the model consumers to adapt the model for the actual deployment environment. We refer to this stage as the deployment stage, to differentiate it from the training and tuning stages occurring at the data controller side [16,17]. In the deployment stage, the model consumers typically have no access to the original training data that are mostly private and proprietary [18,19].…”

Section: Introductionmentioning

confidence: 99%

Supervised Robustness-preserving Data-free Neural Network Pruning

Meng¹,

Bai²,

Teo³

et al. 2022

Preprint

View full text Add to dashboard Cite

When deploying pre-trained neural network models in real-world applications, model consumers often encounter resource-constraint platforms such as mobile and smart devices. They typically use the pruning technique to reduce the size and complexity of the model, generating a lighter one with less resource consumption. Nonetheless, most existing pruning methods are proposed with a premise that the model after being pruned has a chance to be fine-tuned or even retrained based on the original training data. This may be unrealistic in practice, as the data controllers are often reluctant to provide their model consumers with the original data. In this work, we study the neural network pruning in the data-free context, aiming to yield lightweight models that are not only accurate in prediction but also robust against undesired inputs in open-world deployments. Considering the absence of the fine-tuning and retraining that can fix the mis-pruned units, we replace the traditional aggressive one-shot strategy with a conservative one that treats the pruning as a progressive process. We propose a pruning method based on stochastic optimization that uses robustness-related metrics to guide the pruning process. Our method is implemented as a Python package named PAODING and evaluated with a series of experiments on diverse neural network models. The experimental results show that it significantly outperforms existing one-shot data-free pruning approaches in terms of robustness preservation and accuracy.Preprint. Under review.

show abstract

Landscape of Requirements Engineering for Machine Learning-based AI Systems

Cited by 4 publications

References 32 publications

Untitled

Untitled

Brainwriting na elicitação de requisitos para IA confiável

Supervised Robustness-preserving Data-free Neural Network Pruning

Contact Info

Product

Resources

About