On misbehaviour and fault tolerance in machine learning systems

Myllyaho, Lalli; Raatikainen, Mikko; Männistö, Tomi; Nurminen, Jukka K.; Mikkonen, Tommi

doi:10.1016/j.jss.2021.111096

Cited by 19 publications

(6 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In fact, both solutions we have used in the examples of the paper are analogous to patterns of [9] -Oravizio uses the ML model as an Evaluator, and in AuroraAI, User delegation helps to combine data that can only be accessed by the user as a whole. The definition of such patterns remains future work, with some ideas already proposed in [17].…”

Section: Discussionmentioning

confidence: 99%

“…Generalizing this approach implies that it would be possible to build systems so that ML systems are combined, following the pipes-and-filters architectural style [14], with one ML system taking as input the output of another. Unfortunately, as pointed out in our recent study, this does not mean that the models would be immediately composable as such -instead, it is considered preferrable to train a one, single model based on a combined data set than training two models for two different roles [17].…”

Section: B Modelmentioning

confidence: 92%

See 1 more Smart Citation

MLOps Challenges in Multi-Organization Setup: Experiences from Two Real-World Cases

Granlund¹,

Kopponen

Stirbu³

et al. 2021

2021 IEEE/ACM 1st Workshop on AI Engineering - Software Engineering for AI (WAIN)

Self Cite

View full text Add to dashboard Cite

The emerging age of connected, digital world means that there are tons of data, distributed to various organizations and their databases. Since this data can be confidential in nature, it cannot always be openly shared in seek of artificial intelligence (AI) and machine learning (ML) solutions. Instead, we need integration mechanisms, analogous to integration patterns in information systems, to create multi-organization AI/ML systems. In this paper, we present two real-world cases. First, we study integration between two organizations in detail. Second, we address scaling of AI/ML to multi-organization context. The setup we assume is that of continuous deployment, often referred to DevOps in software development. When also ML components are deployed in a similar fashion, term MLOps is used. Towards the end of the paper, we list the main observations and draw some final conclusions. Finally, we propose some directions for future work.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: B Modelmentioning

confidence: 92%

MLOps Challenges in Multi-Organization Setup: Experiences from Two Real-World Cases

Granlund¹,

Kopponen

Stirbu³

et al. 2021

2021 IEEE/ACM 1st Workshop on AI Engineering - Software Engineering for AI (WAIN)

Self Cite

View full text Add to dashboard Cite

show abstract

“…Reproducibility is essential also in the context of fault tolerance, iterative refinement, debugging and optimization of adaptable models, especially for large scale and distributed workflow applications, like cloud computing platforms and Industry 4.0 [32]. The need for specific fault tolerance features for the properties of DL algorithms and their implementations has been already discussed elsewhere [2,36,31]. Reproducibility is also the basis for developing comparison criteria and metrics for the objective evaluation of model properties, like robustness and trustworthiness [33,19].…”

Section: Discussionmentioning

confidence: 99%

Deep Learning Reproducibility and Explainable AI (XAI)

Leventi-Peetz¹,

T.²

2022

Preprint

View full text Add to dashboard Cite

The nondeterminism of Deep Learning (DL) training algorithms and its influence on the explainability of neural network (NN) models are investigated in this work with the help of image classification examples. To discuss the issue, two convolutional neural networks (CNN) have been trained and their results compared. The comparison serves the exploration of the feasibility of creating deterministic, robust DL models and deterministic explainable artificial intelligence (XAI) in practice. Successes and limitation of all here carried out efforts are described in detail. The source code of the attained deterministic models has been listed in this work. Reproducibility is indexed as a development-phase-component of the Model Governance Framework, proposed by the EU within their excellence in AI approach. Furthermore, reproducibility is a requirement for establishing causality for the interpretation of model results and building of trust towards the overwhelming expansion of AI systems applications. Problems that have to be solved on the way to reproducibility and ways to deal with some of them, are examined in this work.

show abstract

“…First, our solution aggregates the scores of Algorithm 1 and Algorithm 2 for a given issue r 0 (Line 1 in Algorithm 3). This kind of aggregation of results of different algorithmsor "voting" -is a useful tool for improving the accuracy of results and masking errors made by one of the algorithms, especially when the components base their results on different metrics [26]. For example, in our case, a direct reference detected by Algorithm 1 to r 0 together with a high similarity score detected by Algorithm 2 implies that someone has noticed the similarity and mentioned it in the comments.…”

Section: Dependency Management Techniquesmentioning

confidence: 99%

“…For the remaining proposals, our solution applies two specific contextualizations (modifications of output checker in [26]) developed based on the feedback of TQC's Jira users: an issue graph-based contextualization and a property-based contextualization. The former emphasizes the dependencies not already in proximity in the issue graph.…”

Section: Dependency Management Techniquesmentioning

confidence: 99%

Improved Management of Issue Dependencies in Issue Trackers of Large Collaborative Projects

Raatikainen

Motger

Lüders

et al. 2023

IIEEE Trans. Software Eng.

View full text Add to dashboard Cite

Issue trackers, such as Jira, have become the prevalent collaborative tools in software engineering for managing issues, such as requirements, development tasks, and software bugs. However, issue trackers inherently focus on the lifecycle of single issues, although issues have and express dependencies on other issues that constitute issue dependency networks in large complex collaborative projects. The objective of this study is to develop supportive solutions for the improved management of dependent issues in an issue tracker. This study follows the Design Science methodology, consisting of eliciting drawbacks and constructing and evaluating a solution and system. The study was carried out in the context of The Qt Company's Jira, which exemplifies an actively used, almost two-decade-old issue tracker with over 100,000 issues. The drawbacks capture how users operate with issue trackers to handle issue information in large, collaborative, and long-lived projects. The basis of the solution is to keep issues and dependencies as separate objects and automatically construct an issue graph. Dependency detections complement the issue graph by proposing missing dependencies, while consistency checks and diagnoses identify conflicting issue priorities and release assignments. Jira's plugin and service-based system architecture realize the functional and quality concerns of the system implementation. We show how to adopt the intelligent supporting techniques of an issue tracker in a complex use context and a large data-set. The solution considers an integrated and holistic system view, practical applicability and utility, and the practical characteristics of issue data, such as inherent incompleteness.

show abstract

On misbehaviour and fault tolerance in machine learning systems

Cited by 19 publications

References 15 publications

MLOps Challenges in Multi-Organization Setup: Experiences from Two Real-World Cases

MLOps Challenges in Multi-Organization Setup: Experiences from Two Real-World Cases

Deep Learning Reproducibility and Explainable AI (XAI)

Improved Management of Issue Dependencies in Issue Trackers of Large Collaborative Projects

Contact Info

Product

Resources

About