About this Research Topic
During the last decade, deep learning techniques, such as convolutional neural networks and long short-term memory, have played a crucial role in the development of the field of computational intelligence. Furthermore, recent research is accelerating the development of sophisticated learning methods in deep neural network architectures and computational environments, typified as the parallel computing technique, for supporting the heavy calculation process required for the adjustment of the large number of network parameters. These benefits of the advancements in deep neural networks are contributing to more emerging applications in various branches of computational intelligence.
Recently, reservoir computing has become the focus of research attention. Reservoir computing consists of a recurrent neural network with a reservoir layer and a readout layer. By utilizing nonlinear spatial-temporal patterns against input signals, the parameters of the readout are learned, thus fixing the synaptic weight of the reservoir layer. The size of the learned parameter in this neural network architecture is significantly lower when compared with that of the other neural network architectures. That is, high learning efficiency is achieved. However, the performance of the conventional network architecture of reservoir computing cannot reach that of the above-mentioned network used for the deep learning approach. Although the conventional reservoir computing network architecture consisted of a single reservoir layer, a recent proposal has been made to modify this to a deep-layer architecture consisting of assembly sub-reservoirs (parallel, ring, hub, and multi-layer types). This enhances the performance of reservoir computing, enabling it to reach the performance of deep learning, along with the maintenance of high learning efficiency, which is an advantage of reservoir computing. This development might, thus, open a new avenue for the application of deep neural network architecture in the field of computational intelligence.
Therefore, our proposed Research Topic is in line with the current trend in the field of deep neural network research, involving deep learning and deep reservoir computing. We welcome Original Research Articles, Brief Research Reports, Reviews, and Mini-reviews addressing, but not limited to, the following issues:
1. Novel deep learning architectures and their applications
• Proposals for novel architectures for deep learning
• Applications to engineering fields, such as bio-medical, controlling, imaging processing, and other fields.
2. Deep reservoir computing architectures and their applications
• Proposals for novel architectures for reservoir computing in mathematical modeling and physical implementation.
• Applications to engineering fields, such as bio-medical, controlling, imaging processing, and other fields.
Manuscript submissions through the Architecture and Systems section should be specifically related to this field and not applications.
Recently, reservoir computing has become the focus of research attention. Reservoir computing consists of a recurrent neural network with a reservoir layer and a readout layer. By utilizing nonlinear spatial-temporal patterns against input signals, the parameters of the readout are learned, thus fixing the synaptic weight of the reservoir layer. The size of the learned parameter in this neural network architecture is significantly lower when compared with that of the other neural network architectures. That is, high learning efficiency is achieved. However, the performance of the conventional network architecture of reservoir computing cannot reach that of the above-mentioned network used for the deep learning approach. Although the conventional reservoir computing network architecture consisted of a single reservoir layer, a recent proposal has been made to modify this to a deep-layer architecture consisting of assembly sub-reservoirs (parallel, ring, hub, and multi-layer types). This enhances the performance of reservoir computing, enabling it to reach the performance of deep learning, along with the maintenance of high learning efficiency, which is an advantage of reservoir computing. This development might, thus, open a new avenue for the application of deep neural network architecture in the field of computational intelligence.
Therefore, our proposed Research Topic is in line with the current trend in the field of deep neural network research, involving deep learning and deep reservoir computing. We welcome Original Research Articles, Brief Research Reports, Reviews, and Mini-reviews addressing, but not limited to, the following issues:
1. Novel deep learning architectures and their applications
• Proposals for novel architectures for deep learning
• Applications to engineering fields, such as bio-medical, controlling, imaging processing, and other fields.
2. Deep reservoir computing architectures and their applications
• Proposals for novel architectures for reservoir computing in mathematical modeling and physical implementation.
• Applications to engineering fields, such as bio-medical, controlling, imaging processing, and other fields.
Manuscript submissions through the Architecture and Systems section should be specifically related to this field and not applications.
Keywords: deep learning, reservoir computing, deep neural network, computational intelligence, engineering applications
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.
