posted by organizer: kcchen || 1665 views || tracked by 2 users: [display]

IEEE-JETCAS-AICHIP 2020 : IEEE JETCAS: Special Issue on Communication-aware Designs and Methodologies for Reliable and Adaptable On-Chip AI SubSystems and Accelerators


When N/A
Where N/A
Submission Deadline May 20, 2020
Notification Due Jul 30, 2020
Final Version Due Aug 15, 2020
Categories    neural network   circuits and systems   memory   on-chip communication

Call For Papers


Call for paper

Communication-aware Designs and Methodologies for Reliable and Adaptable On-Chip AI SubSystems and Accelerators
- Kun-Chih (Jimmy) Chen, National Sun Yat-sen University, Taiwan
- Masoumeh (Azin) Ebrahimi, KTH Royal Institute of Technology, Sweden
- Maurizio Palesi, University of Catania, Italy
- Tim Kogel, Synopsys, Germany
- Manuscript submissions due: 20 May 2020
- Notification of acceptance: 30 July 2020
- Final manuscripts due: 15 August 2020
Go to

The Big data and Internet of Things (IoTs) trend helps to drive the progress of artificial intelligence (AI) technologies in recent years. The notable benefits of neural network (NN) technologies (e.g., artificial neural network (ANN), spiking neural network (SNN), etc.) are widely applied to many real-world applications, such as speech recognition and image classification, and the resulting accuracies have been well above human-level. Due to its undoubted significance, research works of “AI accelerator/subsystem designs” have drawn lots of attention from both academia and industry.

Due to the massive parallel processing, the performance of the current large-scale artificial neural network is often limited by the massive communication overheads and storage requirements. However, the issues of interconnection, communication, computation synchronized with memory subsystem, reliability, and flexibility of AI engines receive less attention. As a result, flexible interconnections and efficient data movement mechanisms for future on-chip AI accelerator are worthy of study, such as:

– New efficient data movement in contemporary AI subsystems: Memory access latency and overhead of NN connections already become the performance bottleneck. The memory area already dominates the total silicon cost, and the data access contributes to a large portion of power consumption in accelerators of deep neural networks (DNN). Many researchers tackled this issue by improving on conventional Von Neumann-type architectures, such as pruning and scheduling. In recent years, novel Non-Von Neumann architectures, such as In-memory Computing and Near-memory Computing techniques, had been proposed extensively to accommodate the computing/processing-in-memory issue, which is worthy of investigation in this special issue.

– Design methodology considering tradeoffs among computing engine (PE arrays) and data movement/storage unit (memory hierarchy) from energy/power/timing point of view: In conventional AI subsystem designs, most works focus on task scheduling to improve the efficiency of the data movement between the memory and the computing engine. In the emerging AI-on-Chip designs, in addition to timing efficiency, power and energy efficiencies become major concerns. Consequently, novel and efficient task scheduling and data movement methodologies from energy/power/timing optimization points of view are covered in this special issue as well.

– “Flexible” and “reliable” communication-aware AI subsystems for future on-chip adaptive learning application: With predefined functional datapath of dedicated neural network models, the computing flows of the current AI accelerators are usually “fixed” and “non-adaptive.” On the other hand, the reliable issue of the SNN design becomes even severe in leading technology nodes because the spike signals are noise sensitive. Therefore, this special issue wants to invite research works of novel computing flows, in consideration of flexible and reliable data movements and communications, for versatile modern AI applications.
Topics of Interest
The special issue solicits high-quality and original contributions on topics including, but not limited to:

– Challenges of massive memory data access in deep learning
– Near- and In-memory computing techniques for saving data movement
– ANN design based on emerging non-volatile memory devices
– Data movement optimization through task scheduling of the artificial neural network
– Energy- and accuracy-aware pruning and quantization mechanism of neural networks
– Novel interconnection networks for the neural networks (e.g., DNN, RNN, ANN, SNN, etc.)
– Efficient on-chip communication of multicore-based artificial neural network
– Communication/traffic-aware artificial neural network architecture and application
– New topology of on-chip communication for efficient neural network computing
– NoC design for heterogeneous ANN computing
– Reliable and robust computing method and on-chip interconnection for AI computing
– Cross-layer optimization for artificial neural network architecture and application
– Tradeoffs among computing engine and data movement from energy/power/timing point of view
– Flexible computing flow and reconfigurable neural network architecture for on-chip learning applications and Artificial General Intelligence (AGI)
– AI-on-Chip interconnections/designs to meet with the requirements and constraints of emerging application (e.g., 5G, medical applications, Industry 4.0, etc.)
– “Flexible” and “reliable” communication-aware AI subsystems for future on-chip adaptive learning application

Related Resources

IEEE/ACM ASONAM 2023   The 2023 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
MIGRATION 2023   MIGRATION, ADAPTATION AND MEMORY - 6th International Interdisciplinary Conference
IEEE Blockchain 2023   The 6th IEEE International Conference on Blockchain
EMOTIONS 2023   MEMORY, AFFECTS AND EMOTIONS - International Interdisciplinary Conference (ONLINE)
IEEE ISCMI 2023   IEEE--2023 10th International Conference on Soft Computing & Machine Intelligence (ISCMI 2023)--SCIE
IEEE ICCC 2023   IEEE--2023 9th International Conference on Computer and Communications (ICCC 2023)
ICSE 2024   The IEEE/ACM International Conference on Software Engineering (Second Round)
TNNLS-GL 2023   IEEE Transactions on Neural Networks and Learning Systems Special Issue on Graph Learning
ICSC 2024   IEEE International Conference on Semantic Computing
IEEE ICSRS 2023   IEEE--2023 The 7th International Conference on System Reliability and Safety (ICSRS 2023)