Current trends and outlook
As a researcher, I work as a Program Committee (PC) member for conferences such as WWW'2018, ISWC'2018, ESWC'2017/2018, and ESWC SemDeep'2018 international workshops. Apart from these, I am also a guest editor for International Semantic Web Journal, Journal of Cloud Computing, and Briefings in Bioinformatics.
While reviewing numerous papers for these conferences and journals, I found that researchers have not limited themselves to developing emerging use cases and analytical solutions using original RNN, CNN, DBN or autoencoders. They are coming up with ideas across new architectures by combining them for diverse domains.
Current trends
As discussed in Chapter 1, Getting Started with Deep Learning, researchers have recently proposed so many emergent DL architectures. These include not only improving CNN/RNN and their variants but also some other special types of architecture: Deep SpatioTemporal Neural Networks (DST-NNs), Multi-Dimensional Recurrent Neural Networks...