Noise Is Useful: Exploiting Data Diversity for Edge Intelligence

Zeng, Zhi; Liu, Yuan; Tang, Weijun; Chen, Fangjiong

doi:10.1109/lwc.2021.3051688

Cited by 11 publications

(3 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Though we used noisy data collected by the real CPSs in the experiments, systematically investigating the impact of noise was not in the scope of our work. Nevertheless, as many studies have already considered the noise issue in machine learning [14,57], they could better guide how to address noisy FOT logs in ENVI.…”

Section: Discussionmentioning

confidence: 99%

Virtual Environment Model Generation for CPS Goal Verification using Imitation Learning

Shin,

Bae

2024

ACM Trans. Embed. Comput. Syst.

View full text Add to dashboard Cite

Cyber-Physical Systems (CPS) continuously interact with their physical environments through embedded software controllers that observe the environments and determine actions. Field Operational Tests (FOT) are essential to verify to what extent the CPS under analysis can achieve certain CPS goals, such as satisfying the safety and performance requirements, while interacting with the real operational environment. However, performing many FOTs to obtain statistically significant verification results is challenging due to its high cost and risk in practice. Simulation-based verification can be an alternative to address the challenge, but it still requires an accurate virtual environment model that can replace the real environment interacting with the CPS in a closed loop. In this paper, we propose ENVI (ENVironment Imitation), a novel approach to automatically generate an accurate virtual environment model, enabling efficient and accurate simulation-based CPS goal verification in practice.To do this, we first formally define the problem of the virtual environment model generation and solve it by leveraging Imitation Learning (IL), which has been actively studied in machine learning to learn complex behaviors from expert demonstrations. The key idea behind the model generation is to leverage IL for training a model that imitates the interactions between the CPS controller and its real environment as recorded in (possibly very small) FOT logs. We then statistically verify the goal achievement of the CPS by simulating it with the generated model. We empirically evaluate ENVI by applying it to the verification of two popular autonomous driving assistant systems. The results show that ENVI can reduce the cost of CPS goal verification while maintaining its accuracy by generating accurate environment models from only a few FOT logs. The use of IL in virtual environment model generation opens new research directions, further discussed at the end of the paper.

show abstract

Section: Discussionmentioning

confidence: 99%

Virtual Environment Model Generation for CPS Goal Verification using Imitation Learning

Shin,

Bae

2024

ACM Trans. Embed. Comput. Syst.

View full text Add to dashboard Cite

show abstract

“…In our case study, though we used noisy data collected by the real CPS, systematically investigating the impact of noise was not in the scope of our work. Nevertheless, as many studies have already considered the noise issue in machine learning [40][41][42], they could better guide how to address noisy data.…”

Section: Discussionmentioning

confidence: 99%

Environment Imitation: Data-Driven Environment Model Generation Using Imitation Learning for Efficient CPS Goal Verification

Shin¹,

Shin²,

Bae³

2022

Preprint

View full text Add to dashboard Cite

Cyber-Physical Systems (CPS) continuously interact with their physical environments through software controllers that observe the environments and determine actions. Engineers can verify to what extent the CPS under analysis can achieve given goals by analyzing its Field Operational Test (FOT) logs. However, it is challenging to repeat many FOTs to obtain statistically significant results due to its cost and risk in practice. To address this challenge, simulation-based verification can be a good alternative for efficient CPS goal verification, but it requires an accurate virtual environment model that can replace the real environment that interacts with the CPS in a closed loop. This paper proposes a novel data-driven approach that automatically generates the virtual environment model from a small amount of FOT logs. We formally define the environment model generation problem and solve it using Imitation Learning (IL) algorithms.In addition, we propose three specific use cases of our approach in the evolutionary CPS development. To validate our approach, we conduct a case study using a simplified autonomous vehicle with a lane-keeping system. The case study results show that our approach can generate accurate virtual environment models for CPS goal verification at a low cost through simulations.

show abstract

“…Farooq et al [77] proposed a federated learning model using Long Short-term Memory (LSTM) neural networks to predict flood that outperforms traditional LSTM models. Furthermore, numerous studies and experiments show that adding noise is an effective mechanism for improving the generalization and robustness of deep neural networks [78][79][80]. This is especially beneficial in imbalanced learning, where new data instances are synthetically generated, and having a training algorithm that can guarantee robustness is crucial.…”

Section: Deep Imbalanced Learningmentioning

confidence: 99%

Federated Learning with GANs-based Synthetic Minority Over-sampling Technique for Improving Weather Prediction from Imbalanced Data

Jafarigol

Trafali̇s

2023

Preprint

View full text Add to dashboard Cite

Detecting rare events is a challenging task among machine learning practitioners, motivating them to navigate and further improve data processing and algorithmic approaches to find accurate and computationally efficient methods for imbalanced learning. Imbalanced data is common in weather prediction, where the massive size of data poses storage and computational challenges. To learn from imbalanced data, the algorithms must strive to learn each class precisely to be able to classify both the minority and the majority class. While creating complex models is somewhat necessary for such problems, the machine learning algorithm must have adequate generalization capabilities. In recent years, federated learning has been used as an environmentally-friendly approach, which can produce accurate results in a distributed setting. Although privacy is not a concern of this study, inherent characteristics of federated learning are beneficial in weather prediction. Deep learning combined with data augmentation is explored in the framework of federated learning. We compare multiple data augmentation methods in a centralized and federated learning framework. Federated learning has great potential for weather prediction tasks. By leveraging data from multiple sources, we were able to improve the accuracy and generalization of the classifier. Addressing the issue of imbalanced data is an essential step. Incorporating the two remarkable approaches of Generative Adversarial Networks and Synthetic Minority Over-sampling Technique is a suitable solution for tabular weather data, and our experimental results confirm the effectiveness of this approach.

show abstract

Noise Is Useful: Exploiting Data Diversity for Edge Intelligence

Cited by 11 publications

References 10 publications

Virtual Environment Model Generation for CPS Goal Verification using Imitation Learning

Virtual Environment Model Generation for CPS Goal Verification using Imitation Learning

Environment Imitation: Data-Driven Environment Model Generation Using Imitation Learning for Efficient CPS Goal Verification

Federated Learning with GANs-based Synthetic Minority Over-sampling Technique for Improving Weather Prediction from Imbalanced Data

Contact Info

Product

Resources

About