Filters








1,141 Hits in 4.4 sec

Automated Safety Verification of Programs Invoking Neural Networks [chapter]

Maria Christakis, Hasan Ferit Eniser, Holger Hermanns, Jörg Hoffmann, Yugesh Kothari, Jianlin Li, Jorge A. Navas, Valentin Wüstholz
2021 Lecture Notes in Computer Science  
In this paper, we embark on the verification of system-level properties for systems characterized by interaction between programs and neural networks.  ...  This shortcoming is pinpointed by programs invoking neural networks despite their acclaimed role as innovation drivers across many application areas.  ...  This work has been supported by DFG Grant 389792660 as part of TRR 248 (see https://perspicuous-computing.science). Jorge Navas has been supported by NSF Grant 1816936.  ... 
doi:10.1007/978-3-030-81685-8_9 fatcat:gs2oyh54zngqrnf6krw3kva5ui

A Review of Formal Methods applied to Machine Learning [article]

Caterina Urban, Antoine Miné
2021 arXiv   pre-print
The large majority of them verify trained neural networks and employ either SMT, optimization, or abstract interpretation techniques.  ...  This raises the question of their safety and their verification. Yet, established formal methods are limited to classic, i.e. non machine-learned software.  ...  Another SAT-based approach for safety verification of binarized neural networks was concurrently proposed by Cheng et al. [30] .  ... 
arXiv:2104.02466v2 fatcat:6ghs5huoynbc5h7lndajmsoxyu

Hierarchical Program-Triggered Reinforcement Learning Agents For Automated Driving [article]

Briti Gangopadhyay, Harshit Soora, Pallab Dasgupta
2021 arXiv   pre-print
The focus of verification shifts to the master program under simple guarantees from the RL agents, leading to a significantly more interpretable and verifiable implementation as compared to a complex RL  ...  Instead of relying on RL agents to learn complex tasks, we propose HPRL - Hierarchical Program-triggered Reinforcement Learning, which uses a hierarchy consisting of a structured program along with multiple  ...  The authors also wish to study how safety specifications can directly be embedded as a part of the DRL agents, thereby eliminating the requirement of an explicit safety shield.  ... 
arXiv:2103.13861v1 fatcat:qtmmz4gmgvetbc2ra3zmijt4oy

ARCH-COMP20 Category Report: Artificial Intelligence and Neural Network Control Systems (AINNCS) for Continuous and Hybrid Systems Plants

Taylor T. Johnson, Diego Manzanas Lopez, Patrick Musau, Hoang-Dung Tran, Elena Botoeva, Francesco Leofante, Amir Maleki, Chelsea Sidrane, Jiameng Fan, Chao Huang
2020 International Workshop on Applied Verification of Continuous and Hybrid Systems  
the most complete assessment of current tools for safety verification of NNCS.  ...  of systems classically known as intelligent control systems, or in more modern and specific terms, neural network control systems (NNCS).  ...  VenMAS VenMAS [2] is a tool for verification of closed-loop systems with neural network components.  ... 
doi:10.29007/9xgv dblp:conf/arch/JohnsonLMTBLMSF20 fatcat:xvlyrekv4nbktcq67nzikys34u

Specifying and Testing k-Safety Properties for Machine-Learning Models [article]

Maria Christakis, Hasan Ferit Eniser, Jörg Hoffmann, Adish Singla, Valentin Wüstholz
2022 arXiv   pre-print
Here, we show the wide applicability of k-safety properties for machine-learning models and present the first specification language for expressing them.  ...  Considering a credit-screening model of a bank, the expected property that "if a person is denied a loan and their income decreases, they should still be denied the loan" is a 2-safety property.  ...  We trained a fully connected neural network of 82% accuracy on the encoded hotel reviews. • Action policies.  ... 
arXiv:2206.06054v1 fatcat:n5r3iw25xzeq3bgxagmihsyqyu

On the Importance of System Testing for Assuring Safety of AI Systems

Franz Wotawa
2019 International Joint Conference on Artificial Intelligence  
Rigorous testing of automated and autonomous systems is inevitable especially in case of safetycritical systems like cars or airplanes.  ...  There exist several functional safety standards that have to be fulfilled like IEC 61508 explicitly stating that AI methodologies are not recommended to be used in case of systems with higher safety requirements  ...  al., 2018c] applying different well-known testing techniques, like mutation testing, combinatorial testing or whitebox testing approaches to neural networks.  ... 
dblp:conf/ijcai/Wotawa19 fatcat:jnwjgcmjgrc4pf74bdkt2xaoau

Safety Verification of Deep Neural Networks [article]

Xiaowei Huang and Marta Kwiatkowska and Sen Wang and Min Wu
2017 arXiv   pre-print
We develop a novel automated verification framework for feed-forward multi-layer neural networks based on Satisfiability Modulo Theory (SMT).  ...  by a human, and define safety for an individual decision in terms of invariance of the classification within a small neighbourhood of the original image.  ...  In this paper we propose a general framework for automated verification of safety of classification decisions made by feed-forward deep neural networks.  ... 
arXiv:1610.06940v3 fatcat:qenhqyyr2vbabfyj2ft3gsk2zy

Safety Verification of Deep Neural Networks [chapter]

Xiaowei Huang, Marta Kwiatkowska, Sen Wang, Min Wu
2017 Lecture Notes in Computer Science  
We develop a novel automated verification framework for feed-forward multi-layer neural networks based on Satisfiability Modulo Theory (SMT).  ...  by a human, and define safety for an individual decision in terms of invariance of the classification within a small neighbourhood of the original image.  ...  In this paper we propose a general framework for automated verification of safety of classification decisions made by feed-forward deep neural networks.  ... 
doi:10.1007/978-3-319-63387-9_1 fatcat:te7nwtabjfcgdn3brho5ff4j7i

Certification considerations for adaptive systems

S. Bhattacharyya, D. Cofer, D. Musliner, J. Mueller, E. Engstrom
2015 2015 International Conference on Unmanned Aircraft Systems (ICUAS)  
These aircraft will likely incorporate adaptive control algorithms that will provide enhanced safety, autonomy, and robustness during adverse conditions.  ...  and identifying gaps and challenges associated with certification of each approach.  ...  Siddhartha Bhattacharyya of Rockwell Collins provided general oversight, outline of the sections and characterization.  ... 
doi:10.1109/icuas.2015.7152300 fatcat:hqa4gpy4kvcntoqbcpxbnbwn74

Machine Learning and Artificial Intelligence in Neurosurgery: Status, Prospects, and Challenges

T Forcht Dagi, Fred G Barker, Jacob Glass
2021 Neurosurgery  
ML constitutes a subfield of AI. DL is sometimes classified as a subfield of ML (and therefore of AI) and sometimes of neural networks. Neural networks are the basis for DL algorithms.  ...  ML uses AI, neural networks, and automated algorithms to accomplish its objectives. Data needs Relative to ML, DM can produce results on lesser volumes of data.  ...  Disclosures The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article.  ... 
doi:10.1093/neuros/nyab170 pmid:34015816 fatcat:pviguom63nfjlepk2ydsrtewde

AMASS: Automated Software Mass Customization via Feature Identification and Tailoring

Hongfa Xue, Yurong Chen, Guru Venkataramani, Tian Lan
2018 EAI Endorsed Transactions on Security and Safety  
accuracy for feature/function identification and up to 67% reduction of program attack surface.  ...  The rapid inflation of software features brings inefficiency and vulnerabilities into programs, resulting in an increased attack surface with a higher possibility of exploitation.  ...  Any opinions, findings, conclusions, or recommendations expressed in this article are those of the authors, and do not necessarily reflect those of ONR.  ... 
doi:10.4108/eai.13-7-2018.162291 fatcat:3pywewt7lzgmvn7z4qppqhpfoi

Towards Scalable Verification of Deep Reinforcement Learning [article]

Guy Amir, Michael Schapira, Guy Katz
2021 arXiv   pre-print
Deep neural networks (DNNs) have gained significant popularity in recent years, becoming the state of the art in a variety of domains.  ...  To demonstrate the benefits of whiRL 2.0, we apply it to case studies from the communication networks domain that have recently been used to motivate formal verification of DRL systems, and which exhibit  ...  Israel Science Foundation (grant number 683/18), the Binational Science Foundation (grant numbers 2017662 and 2019798), and the Center for Interdisciplinary Data Science Research at The Hebrew University of  ... 
arXiv:2105.11931v2 fatcat:whbvemzqjnckfd6tjalefev5hu

Introduction to Neural Network Verification

Aws Albarghouthi
2021 Foundations and Trends® in Programming Languages  
In many settings, we need to provide formal guarantees on the safety, security, correctness, or robustness of neural networks.  ...  This monograph covers foundational ideas from formal verification and their adaptation to reasoning about neural networks and deep learning.  ...  Acknowledgements References Full text available at: http://dx.doi.org/10.1561/2500000051 Acknowledgements Thanks to the best focus group ever: the CS 839 students and TA, Swati Anand, at the University of  ... 
doi:10.1561/2500000051 fatcat:2wbm374jcrc3pd5mhn5pfukoae

Climate Modeling System with Adaptation of Neural Network and AI Data Mining Techniques

Ahmed Mateen, Salman Afsar, Amir Waheed, Zulfiqar Ali
2016 International Journal of Computer Applications  
In this paper, data mining procedures are used with generalized Neural Network technique which is useful for weather forecasting quickly with the help of data clustering and screening.  ...  The purpose of this study is to develop a climate modeling system by using data mining techniques which are the process of extracting needed information's from the large database.  ...  Therefore, by creating a neural network model for both short and long-term forecasts outperforms the other models.  ... 
doi:10.5120/ijca2016911908 fatcat:h3o3by3gafdbpo7jfodoma74x4

ALLSTAR: A Blockchain Based Decentralized Ecosystem for Cloud and Edge Computing

Huan Zhou, Xue Ouyang, Zhiming Zhao
2020 2020 IEEE International Conference on Joint Cloud Computing  
However, the centralized management mechanism of current Clouds lacks the dispersion to satisfy the requirements of emerging collaborative applications, including AI, IoT, and autopilot.  ...  This paper describes the overall architecture of ALLSTAR, the related key techniques, and detailed application DevOps processes as well as the new business model.  ...  It is also supported by the National Key Research and Development Program of China (2016YFB1000100). The authors would also like to thank Alexandre from MOG technologies providing the use case.  ... 
doi:10.1109/jcc49151.2020.00018 fatcat:xfgvnuxrhreezbslqhvcqmsujm
« Previous Showing results 1 — 15 out of 1,141 results