About this Research Topic
Deep learning has considerably facilitated feature engineering and enabled deep knowledge excavation for mining unstructured and high-dimensional data. However, the advancement of deep learning methods is accompanied by significantly increased model complexity. The incurred computational overhead negatively affects the efficiency when it comes to large-scale datasets or resource-constrained scenarios. Given the prevalence of mining big data in real-time and mining massive data streams, the efficiency of deep learning methods for big data mining has been a serious concern for real-world big data mining applications.
Recently, many deep learning models have shown noticeable potential for efficiency improvement without performance degradation, which has a noteworthy influence on big-data mining. For instance, in many computer vision tasks, the very widely used convolutional neural network has demonstrated its tolerability to dozens of times compression ratio, where the performance is even enhanced. Though similar results have been extensively observed in newly proposed deep learning models (e.g., deep generative models, graph neural networks, neural ordinary differential equations, Transformers), a systematic investigation on efficiency improvement in the context of big data mining is demanded to be conducted for advancing both areas. Furthermore, recent trends in multimodal data mining, interactive data mining, and incremental data mining pose eager needs for efficient data processing. The goal of this research topic is to bring together theories and applications of efficient deep learning techniques to big-data mining problems. The proposed research theme will focus on efficient deep learning techniques for big data mining.
The topics of interest include but are not limited to the following areas:
• Neural Network Pruning
• Sparse Deep Learning
• Model Quantization
• Knowledge Distillation
• Automated lightweight model design
• Automated deep learning acceleration
• Model compression for real-time data mining
• Efficient algorithms for streaming data mining
• Efficient deep model for interactive big-data mining
The Research Topic Coordinator is Dr. Yao Zhou, who is from Sichuan University, and his current research interests include model compression and evolutionary computation.
Recently, many deep learning models have shown noticeable potential for efficiency improvement without performance degradation, which has a noteworthy influence on big-data mining. For instance, in many computer vision tasks, the very widely used convolutional neural network has demonstrated its tolerability to dozens of times compression ratio, where the performance is even enhanced. Though similar results have been extensively observed in newly proposed deep learning models (e.g., deep generative models, graph neural networks, neural ordinary differential equations, Transformers), a systematic investigation on efficiency improvement in the context of big data mining is demanded to be conducted for advancing both areas. Furthermore, recent trends in multimodal data mining, interactive data mining, and incremental data mining pose eager needs for efficient data processing. The goal of this research topic is to bring together theories and applications of efficient deep learning techniques to big-data mining problems. The proposed research theme will focus on efficient deep learning techniques for big data mining.
The topics of interest include but are not limited to the following areas:
• Neural Network Pruning
• Sparse Deep Learning
• Model Quantization
• Knowledge Distillation
• Automated lightweight model design
• Automated deep learning acceleration
• Model compression for real-time data mining
• Efficient algorithms for streaming data mining
• Efficient deep model for interactive big-data mining
The Research Topic Coordinator is Dr. Yao Zhou, who is from Sichuan University, and his current research interests include model compression and evolutionary computation.
Keywords: deep learning, model compression, knowledge distillation, streaming data mining, interactive big data mining
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.
