全文截稿: 2021-06-30
影响因子: 1.045
CCF分类: C类
中科院JCR分区:
• 大类 : 工程技术 - 4区
• 小类 : 计算机：硬件 - 4区
• 小类 : 计算机：理论方法 - 4区
• 小类 : 工程：电子与电气 - 4区
网址：
https://www.journals.elsevier.com/microprocessors-and-microsystems
Deep reinforcement learning (DRL) uses feedback from the agent to make decisions in complex problems under uncertainty. Medical applications often require processing large volumes of complex data in a challenging environment. Deep reinforcement learning can process this data by analyzing the agent's feedback that is sequential and sampled using non-linear functions. The deep reinforcement learning algorithms commonly used for medical applications include value-based methods, policy gradient, and actor-critic methods. The recent advances in the increased computational capabilities of architectures like field-programmable gate array (FPGA), graphics processing units (GPU), and digital signal processors (DSP) have made it possible to infer deep reinforcement learning algorithms on them. However, efficient implementation of these architectures should consider the issues related to their portability, wearability, and power consumption.
The main objective is to provide a platform for scientists, researchers, industry experts, and scholars to share their innovative contributions in deep reinforcement learning for medical applications on various embedded devices (ED). Research articles describing only a proof of concept are not encouraged. Authors are solicited to develop novel deep reinforcement learning algorithms on medical data and implement them either on FPGA, GPUs, or DSP. The special issue invites authors to submit papers that analyze the portability, wearability, power consumption of the deep reinforcement learning algorithms implemented either on FPGA, GPU, or DSP. The deep reinforcement learning topic includes but not restricted to:
Monte Carlo Tree Search and Deep Q-network
Dual Gradient Descent and Conjugate Gradient
Trust Region Policy Optimization and Proximal Policy Optimization.
Actor-Critic using Kronecker-Factored Trust Region
Linear Quadratic Regulator and Iterative Linear Quadratic Regulator
Twin-Delayed Deep deterministic policy gradient
Guided Policy Search and Model-Based Learning with Raw Medical Videos
Inverse Reinforcement Learning and Meta-learning
Very efficient ED for DRL in medical applications in terms of power consumption, processing efficiency and flexibility
Neuromorphic and/or brain-inspired architectures implementing DRL techniques
Efficient mapping of DRL applications to ED
New learning approaches for DRL targeting ED
High-level programming language support, tools, frameworks, and system software for DRL in medical applications implemented on ED
Security and Reliability issues for DRL on ED
DRL ED implementation in cyber-physical systems for healthcare, well-being and personal assistance (elderly, disability), sports and medicine, rehabilitation, instrumentation, lab-on-chips
Important dates
Paper submission due: June 30, 2021
First-round acceptance notification: August 30, 2021
Revision submission: October 15, 2021
Notification of final decision: December 30, 2021
Submission of final paper: January 30, 2022
Publication date: March 2022
