Thank dear Adnan Riaz ,Muhammad Ali and Reema Ahmad a lot for your kind and Insightful reply. The original intention of the question is to invite outstanding scholars in related fields to discuss and exchange the general model of sequence learning. Some years ago, when we talk about CV, we may talk about CNN while when we talk about time series (Speech, NLP) , we may talk about RNN (LSTM, GRU) given massive data. Now some of the frontier research in the academic field makes us rethink the internal relationship between the model and the field. We can discuss any specific ideas under this topic in an open and inclusive way, which I think is very interesting, challenging and meaningful for the future of sequence learning.

Yes, you’re right. Now it is not necessary to use RNN in above mentioned domains. Some researchers already propsoed CNN based NN architectures and in some cases CNN based architecture is more better than RNN based networks in performance.

In time series , CNN based architecture is more suitable as comapred to RNN based. There are some hybrid architectures based on CNN and RNN and outperforms in time series

I am agree with Adnan Riaz. Sequential Learning for Longitudinal data is one the recent trend of great focus, see e.g., https://www.frontiersin.org/articles/10.3389/fgene.2019.00617/full

Article The Effects of Family Cohesiveness and Peer Encouragement on...

https://www.cs.cmu.edu/~epxing/Class/10708-17/project-reports/project15.pdf

LSTM (Long Short Term Memory) is the future of general deep learning for sequential learning

Thank dear Adnan Riaz ,Muhammad Ali and Reema Ahmad a lot for your kind and Insightful reply. The original intention of the question is to invite outstanding scholars in related fields to discuss and exchange the general model of sequence learning. Some years ago, when we talk about CV, we may talk about CNN while when we talk about time series (Speech, NLP) , we may talk about RNN (LSTM, GRU) given massive data. Now some of the frontier research in the academic field makes us rethink the internal relationship between the model and the field. We can discuss any specific ideas under this topic in an open and inclusive way, which I think is very interesting, challenging and meaningful for the future of sequence learning.

Despite CNN and RNN, GCN is also helpful in handling many NLP applications for sequence learning incorporated with GANs structures. In my perspective, all deep learning-based networks either CNN, RNN, or GCN have an uniques impact on different applications in this scientific era. Yiwen Sun Reema Ahmad Muhammad Ali Adnan Riaz

Quite interesting and thanks for your idea Muhammad Nabeel . The GCN you mentioned does have a remarkable success in building relationships between objects. In the impression, there are also some works trying to explore the relationship between self attention mechanism and GCN From a theoretical point of view.

you are right Yiwen Sun and furthermore, a hybrid model including attention mechanism + GAN+GCN can be exploited in different applied NLP, AI, and image processing applications as domain-specific systems (Air-ticket recommendation systems, network traffic speed prediction systems, E-healthcare systems), etc.

Yiwen Sun I do agree with Muhammad Nabeel , Look at a recent publication in NLP domain , they used GCN based architecture and results are quite surprising.

Article Cas-GANs: An Approach of Dialogue Policy Learning based on G...

Dear Researcher Reema Ahmad , I am partially disagree with you , because initially it's thought that for Sequential learning/time series only LSTM based models can be used but now the trend has changed even in some time series problems CNN based models outperformed over RNN based models. So , Now it's hard to say that "LSTM (Long Short Term Memory) is the future of general deep learning for sequential learning. "