Deep learning, as a fundamental part of human perception, has promoted the development of cutting-edge robotic systems with the ability to automatically mine concepts from complex tasks in an open-ended manner. Many novel algorithms and efficient architectures of deep learning with trainable components have achieved remarkable performance in various domains such as artificial software and robotic devices, based on the unsupervised/supervised learning schemes.
Most of the current deep learning methods focus on single-view perception of objects without fully considering the intrinsic characteristics of data that objects can be described by heterogeneous views. For example, in the interpretation of multimedia content, the semantics of multimedia segments can be described by different data structures, such as images and audios. In web-page recognition, a web page can be represented by content vectors of texts and relation graphs of hyperlinks. Those heterogeneous views contain complementary knowledge and information that can further improve representation learning of data. With the development for easier access to heterogeneous view data promoted by wider deployments of edge-computing robotic devices, deep heterogeneous view perception distilling knowledge from various views is increasingly attracting more attention. At the same time, heterogeneous view data contains more private information than single view data. It is inevitable that mining large-scale heterogeneous view data will raise the issue of privacy, and with the emergence of deep heterogeneous view perception, privacies hidden in data are becoming more fragile to leak. Thus, to perceive deep heterogeneous view knowledge of data without lacking privacies is also becoming the core of neural computing.
Thus, this Research Topic aims to motivate novel theories and applications of deep heterogenous-view perception driven by new neural computing architectures and hardware in Big Data, as well as the privacy-preserving theories and applications in heterogeneous view perception. The sub-topics include, but are not limited to:
1. Efficient neural computing paradigms accelerated by the quantum neural computing and crowd-source neural computing systems;
2. New neural fusion mechanisms in merging information of heterogenous-view data;
3. New neural learning algorithms for the heterogeneous view perception of big data;
4. Intelligent neural computing hardware accelerating the deep heterogeneous view perception;
5. Fuzzy data representations promoted by tensor and graph computing theories;
6. Privacy-preserving computing methods based on blockchain and fully homomorphic encryption for the heterogeneous view perception;
7. New applications of deep heterogeneous view perception in big data;
All contributions should be directly or indirectly related to the software or hardware of neurorobotics with the aim to potentially boost the development of robotic systems.
Deep learning, as a fundamental part of human perception, has promoted the development of cutting-edge robotic systems with the ability to automatically mine concepts from complex tasks in an open-ended manner. Many novel algorithms and efficient architectures of deep learning with trainable components have achieved remarkable performance in various domains such as artificial software and robotic devices, based on the unsupervised/supervised learning schemes.
Most of the current deep learning methods focus on single-view perception of objects without fully considering the intrinsic characteristics of data that objects can be described by heterogeneous views. For example, in the interpretation of multimedia content, the semantics of multimedia segments can be described by different data structures, such as images and audios. In web-page recognition, a web page can be represented by content vectors of texts and relation graphs of hyperlinks. Those heterogeneous views contain complementary knowledge and information that can further improve representation learning of data. With the development for easier access to heterogeneous view data promoted by wider deployments of edge-computing robotic devices, deep heterogeneous view perception distilling knowledge from various views is increasingly attracting more attention. At the same time, heterogeneous view data contains more private information than single view data. It is inevitable that mining large-scale heterogeneous view data will raise the issue of privacy, and with the emergence of deep heterogeneous view perception, privacies hidden in data are becoming more fragile to leak. Thus, to perceive deep heterogeneous view knowledge of data without lacking privacies is also becoming the core of neural computing.
Thus, this Research Topic aims to motivate novel theories and applications of deep heterogenous-view perception driven by new neural computing architectures and hardware in Big Data, as well as the privacy-preserving theories and applications in heterogeneous view perception. The sub-topics include, but are not limited to:
1. Efficient neural computing paradigms accelerated by the quantum neural computing and crowd-source neural computing systems;
2. New neural fusion mechanisms in merging information of heterogenous-view data;
3. New neural learning algorithms for the heterogeneous view perception of big data;
4. Intelligent neural computing hardware accelerating the deep heterogeneous view perception;
5. Fuzzy data representations promoted by tensor and graph computing theories;
6. Privacy-preserving computing methods based on blockchain and fully homomorphic encryption for the heterogeneous view perception;
7. New applications of deep heterogeneous view perception in big data;
All contributions should be directly or indirectly related to the software or hardware of neurorobotics with the aim to potentially boost the development of robotic systems.
