CALL FOR PAPERS

The PeRConAI workshop aims at fostering the development and circulation of new ideas and research directions on pervasive and resource-constrained machine learning} bringing together practitioners and researchers working on the intersection between pervasive computing and machine learning, stimulating the cross-fertilization between the two communities. The PeRConAI workshop solicits contributions on, but not limited to, the following topics:

• Foundations of Advanced Machine learning algorithms and methods for pervasive systems subject to resource limitations addressing the following open challenges:
  • distributed/decentralized ML/DL for resource-constrained devices (e.g., federated deep learning in centralized and decentralized settings, sharing deep neural networks at the edge, cross-device edge federated deep learning);
  • optimization of distributed/decentralized learning systems in pervasive scenarios (e.g. resource-efficient federated learning with or without central coordinator) and novel algorithms to distribute pervasive deep learning mechanisms and fog computing approaches to support distributed deep neural network applications;
  • parallel and edge computing techniques to support the widespread usage of different deep neural network architectures;
  • the definition/application of lightweight ML/DL models for on-device training/inference in pervasive computing (e.g., GRU, ELM, etc.);
  • compression/pruning of machine learning models for real-time inference (e.g., pruning, quantization, sparsification, lottery ticket hypothesis, knowledge distillation for both training and inference) and distributed management of a large amount of data for deep learning as a service at the edge;
  • privacy-preserving distributed/decentralized ML/DL algorithms and systems in pervasive and resource-constrained scenarios;
  • semi-supervised and self-supervised learning systems in pervasive and resource-constrained scenarios;
  • learning with imbalanced data in pervasive and resource-constrained scenarios and deep learning-based architectures for low-power and limited resources devices;
  • continual learning in pervasive and resource-constrained scenarios;
  • usage of deep learning to improve the performance of current distributed and parallel computing techniques improving tasks or data allocations across smart devices.
• Applications of Advanced Machine learning algorithms, methods and approaches for pervasive computing under resource-limitations applied to the following application domains:
  • Health and well-being applications (e.g. activity recognition, health monitoring, etc.).
  • Anomaly/Novelty detection (e.g. Industry 4.0, intrusion detection, privacy, and security, etc.).
  • Environmental applications (e.g. meteorology, biology, environmental disaster prevention/detection).
  • Audio signal processing (e.g., sound event detection, speech recognition/processing).
  • Video streams processing on resource-constrained devices.
  • Natural Language Processing and Information Retrieval (e.g. conversational applications running on mobile or edge devices).
  • Intersection between mobile computing with ML/DL on resource-constrained devices.
  • Any other real-world applications and case studies where the pervasiveness of resource-constrained devices is central for knowledge extraction.