A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf
.
Filters
Model Assertions for Monitoring and Improving ML Models
[article]
2020
arXiv
pre-print
We propose a new abstraction, model assertions, that adapts the classical use of program assertions as a way to monitor and improve ML models. ...
To prevent errors, ML engineering teams monitor and continuously improve these models. ...
We further acknowledge Kayvon Fatahalian, James Hong, Dan Fu, Will Crichton, Nikos Arechiga, and Sudeep Pillai for their productive discussions on ML applications. ...
arXiv:2003.01668v3
fatcat:bh7csm3skfcqzdis334tpz5fye
The ML test score: A rubric for ML production readiness and technical debt reduction
2017
2017 IEEE International Conference on Big Data (Big Data)
Testing and monitoring are key considerations for ensuring the production-readiness of an ML system, and for reducing technical debt of ML systems. ...
improve production readiness and pay down ML technical debt. ...
ACKNOWLEDGMENT We are very grateful to Keith Arner, Gary Holt, Josh Lovejoy, Fernando Pereira, Todd Phillips, Tal Shaked, Todd Underwood, Martin Wicke, Cory Williams, and Martin Zinkevich for many helpful ...
doi:10.1109/bigdata.2017.8258038
dblp:conf/bigdataconf/BreckCNSS17
fatcat:qjmoshv3grbgxgj3et7vtdfdu4
Unsolved Problems in ML Safety
[article]
2022
arXiv
pre-print
We present four problems ready for research, namely withstanding hazards ("Robustness"), identifying hazards ("Monitoring"), reducing inherent model hazards ("Alignment"), and reducing systemic hazards ...
In response to emerging safety challenges in ML, such as those introduced by recent large-scale models, we provide a new roadmap for ML Safety and refine the technical problems that the field needs to ...
Sculley, Mark Xu, Beth Barnes, Andreas Terzis, Florian Tramèr, Stella Biderman, Leo Gao, Jacob Hilton, and Thomas Dietterich for their feedback. ...
arXiv:2109.13916v4
fatcat:5jq5qxxi3nfuvhboaehhekybfi
Towards safety monitoring of ML-based perception tasks of autonomous systems
2020
2020 IEEE International Symposium on Software Reliability Engineering Workshops (ISSREW)
However, applying an SM for ML components can be complex in terms of detection and reaction. ...
Thus, aiming at dealing with this challenging task, this work presents a benchmark architecture for testing ML components with SM, and the current work for dealing with specific ML threats. ...
The author thanks Jérémie Guiochet, Hélène Waeselynck, Mario Trapp, and Harita Joshi for their guidance during this PhD's 1st year. ...
doi:10.1109/issrew51248.2020.00052
fatcat:exz2ngbkwfcr3aftndthbbokie
Survey on Machine Learning Algorithms Enhancing the Functional Verification Process
2021
Electronics
Current research of deploying different (ML) models prove to be promising in areas such as stimulus constraining, test generation, coverage collection and bug detection and localization. ...
Machine learning (ML) models proved to be valuable for automating major parts of the process, which have typically occupied the bandwidth of engineers; diverting them from adding new coverage metrics to ...
For the process of generating assertions, presented resources managed to extract effective assertions with the use of ML models, such as DT and RNN, from existing designs and from input natural language ...
doi:10.3390/electronics10212688
fatcat:ic2ub7423rcf3exbjatizaoofu
Knowledge-Based Architecture for Recognising Activities of Older People
2019
Procedia Computer Science
Supervised Machine Learning (ML) algorithms are the most commonly used techniques for this application. ...
Intraand inter-personal variation in performing complex activities is another challenge for an ML-based activity recognition approach. ...
Acknowledgements This research was part funded by UK EPSRC grants EP/P01013X/1 (STRETCH) and EP/R013144/1 (SAUSE) and ERC grant 291652 (ASAP). ...
doi:10.1016/j.procs.2019.09.214
fatcat:big3mnxpird75naynjerkmmqky
Data management research at the MITRE Corporation
1995
SIGMOD record
We would also like to thank Ron Haggarty, MITRE's Vice President for Research and Technology, for his committment to research in this area. ...
ACKNOWLEDGMENTS The authors would like to thank Pamela Campbell, Linda Chambless, and Manette Lazear for providing ideas, project descriptions, and valuable feedback. ...
We are investigating MLS support for important OODBMS features that are missing from earlier MLS OODBMS models. ...
doi:10.1145/211990.212020
fatcat:hqb7wxnpenezlbdy6rk4bx3sme
Beyond technology: Can artificial intelligence support clinical decisions in the prediction of sepsis?
Para além da tecnologia: a inteligência artificial pode apoiar decisões clínicas na predição da sepse?
2022
Revista Brasileira de Enfermagem
Para além da tecnologia: a inteligência artificial pode apoiar decisões clínicas na predição da sepse?
Conclusions: Far beyond technology, ML models can speed up assertive clinical decisions by nurses, optimizing time and specialized human resources. ...
The Machine Learning (ML) tool, Robot Laura®, scores changes in vital parameters and lab tests, classifying them by severity. Inpatients and patients over 18 years of age were included. ...
CONCLUSIONS Far beyond technology, ML models can speed up assertive clinical decisions by nurses, through critical alarms, optimizing time and specialized human resources. ...
doi:10.1590/0034-7167-2021-0586
fatcat:nulq5oivpff6rby2s6jsjqowma
Towards Observability for Machine Learning Pipelines
[article]
2022
arXiv
pre-print
We propose a new type of data management system that offers end-to-end observability, or visibility into complex system behavior, for ML pipelines through assisted (1) detection, (2) diagnosis, and (3) ...
and silent failures that could occur at any stage, or component, of the ML pipeline (e.g., data distribution shift). ...
The majority of the work in data management for ML concentrates on specific components, e.g., for identifying data bugs during preprocessing [1, 2] , or for logging models and model metadata for post-hoc ...
arXiv:2108.13557v2
fatcat:ddgupkuskvcf5n4xhkybr5f36i
Benchmarking Safety Monitors for Image Classifiers with Machine Learning
[article]
2021
arXiv
pre-print
As the prediction from the ML is the core information directly impacting safety, many works are focusing on monitoring the ML model itself. ...
Therefore, this paper aims at establishing a baseline framework for benchmarking monitors for ML image classifiers. ...
This technique is an adaptation of the classical program assertions to monitor and improve ML models. The idea is to verify inputs/outputs that indicate when errors may be occurring in the system. ...
arXiv:2110.01232v1
fatcat:t4jj7uh66vhftoiiea5eyajxma
Benchmarking Safety Monitors for Image Classifiers with Machine Learning
2021
2021 IEEE 26th Pacific Rim International Symposium on Dependable Computing (PRDC)
HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. ...
The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. ...
This technique is an adaptation of the classical program assertions to monitor and improve ML models. The idea is to verify inputs/outputs that indicate when errors may be occurring in the system. ...
doi:10.1109/prdc53464.2021.00012
fatcat:wvwbz22iz5egbesdirazuxpqsm
Intravenous Lipid Emulsion Does Not Augment Blood Pressure Recovery in A Rabbit Model of Metoprolol Toxicity
2010
Journal of Medical Toxicology
Effect for ILE has been demonstrated in animal models of propranolol poisoning; however, any benefit for ILE in more hydrophilic β-blockers remains uncertain. ...
Twenty sedated, invasively monitored and mechanically ventilated adult New Zealand white rabbits underwent metoprolol infusion to mean arterial pressure (MAP) 60% baseline. ...
for improvement in the latter. ...
doi:10.1007/s13181-010-0049-y
pmid:20354918
pmcid:PMC3550458
fatcat:rtzx3ghsirdwtkkfwnvdyzhlby
Page 7 of Primary Cardiology Vol. 16, Issue 1
[page]
1990
Primary Cardiology
The investigators assert that their two-dimensional echocardiography studies suggest that LV modeling and function are improved in patients who receive IV strepto- kinase after Mi compared with those assigned ...
End systolic volume was signifi- cantly less (P = 0.036) for streptokinase-treated patients (65.4 + 36.4 mL) than for control patients (74.9 + 45.7 mL). ...
Integration of formal and heuristic reasoning as a basis for testing and debugging computer security policy
1993
Proceedings on the 1992-1993 workshop on New security paradigms - NSPW '92-93
of information systems to these threats, especially for those aspects of information systems that are more readily amenable to modeling via non-formal methods. ...
We present a paradigm integrating formal and heuristic reasoning as a basis for testing for and debugging computer security policy. ...
Acknowledgements Richard Wexelblat actively participated in research meetings which produced some of the ideas which were subsequently further refined and presented in this paper. ...
doi:10.1145/283751.283784
dblp:conf/nspw/MichaelSL93
fatcat:bw7bnqxzgbe2fnlceqxr5rzao4
Improved Tool Support for Machine-Code Decompilation in HOL4
[chapter]
2015
Lecture Notes in Computer Science
As a result of these improvements, decompilation is faster (on average by one to two orders of magnitude), the instruction set specifications are easier to write, and the proof tools are easier to maintain ...
This paper presents improvements that have been made to our methodology for soundly decompiling machine-code programs to functions expressed in HOL logic. ...
To date, L3 models have been exported to HOL4, Standard ML, Isabelle/HOL and TSL [10] . ...
doi:10.1007/978-3-319-22102-1_12
fatcat:xrbtb3mlrzezrddt2zrci2rl54
« Previous
Showing results 1 — 15 out of 19,522 results