Call for paper for the Special issue: Machine Learning for Safety Critical Systems
• 大类 : 管理科学 - 2区
• 小类 : 工程：工业 - 2区
• 小类 : 运筹学与管理科学 - 2区
Aim and Scope
The special issue invites submissions on Machine Learning for Safety Critical Systems – a research discipline where a systems’ reliability and ability to respond in critical situations becomes more prudent. Modern engineering fields such as automotive, aerospace, robotic, and networks, are exploiting machine learning to improve and maintain mission critical activities. These systems are large, complex and often require real-time learning with feedback to ensure they function as desired. Since detecting anomalies, analysing failures and predicting future system states are quickly becoming part of the engineering design process, algorithmic issues when making real-time decisions based on fast arriving, high-volume condition data, on-site feedback and data models has become the focal point of many discussions.
This special issue aims to bring together diverse researchers from areas such as reinforcement learning, autonomous agents, game theory, controls and operations engineering teams to develop approaches which enable real-time discovery, inference and computational tools. These techniques are aimed to influence engineering operations that automate mission-critical and safety applications. The focus is placed on aspects of general machine learning algorithms to solve problems for engineering domains. However, the editors also encourage exploration of new innovative machine learning approaches, which can solve problems with improved latency. We are also seeking contributions in advances of streaming and distributed algorithms, heterogeneous and high dimensional data sets to be used for real-time decision-making for critical safety measures.
Topics of interest
• Adaptation: How can systems learn and adapt to changes in the environment (especially in dynamic environments) when training data is less and requires quick model assumption. How can principles of autonomous agents working together to build large engineering systems be exploited to react in dynamic situations.
• Noisy and poor data sets: How can machine learning models be trained to understand noisy data sets for quick learning. Missing data exploration?
• Detecting anomalous behaviour: How can anomalies be detected quickly and partitioned appropriately such that correct actions are applied?
• Improving latency: How can machine learning algorithms be improved to produce results quickly than previously anticipated?
• Improving software and hardware performance: Exploring models of GPU, HPC processing and FPGAs to improve the performance of algorithms can greatly influence their use in engineering design. Experimental demonstrations are encouraged.
• Reinforcement learning: How can machine learn correct behaviour? Can training be made quicker with guidance to allow algorithms to produce corrective measure when anomalies are detected?
• Human factors: how can engineers maintaining the system interact with the selfautonomous system
• Open problems in engineering where machine learning is not proving fruitful. What are the open problems in operations where practical machine learning is difficult to apply? What are the limitations and how can these be improved?