Hi dear scholars,
As thousand of researches are going on and daily every one proposing new methods on deep neural networks, now there are so many variations, and it is very difficult to follow up. I recently learned that GNN (Graph Neural Network) is getting popularity.
Is the any website or portal where we can follow up those recent developments and read only very important once, not every thing what is published?
Thank you in advance.
https://www.journals.elsevier.com/neural-networks
https://mitpress.mit.edu/books/advances-neural-information-processing-systems
https://www.nature.com/search?q=Graph+Neural+Network
https://journals.plos.org/plosone/search?q=Graph+Neural+Network&filterJournals=PLoSONE
https://scholar.google.es/scholar?hl=es&as_sdt=0%2C5&q=Graph+Neural+Network&btnG=
Dear Ochilbek Rakhmanov
Apart from the Graph neural network, deep convolution neural networks are quite popular and in trend. Here are few relevant links related to recent neural networks:
1. https://www.journals.elsevier.com/neural-networks/recent-articles
2. Article A Survey of the Recent Architectures of Deep Convolutional N...
3. Article Recent Advances in Convolutional Neural Networks
Regards
Ochilbek Rakhmanov
Usually, there are some researchers and/or labs working on a dedicated topic that publish a list of relevant papers and/or available code on github. Besides this list is regularly updated so that you can be informed of the recent work.
Here is one of the most relevant I know so far regarding GNN:
https://github.com/thunlp/GNNPapers
From what I can see, there are more and more "specialized" network architectures, i.e., transformers for instance which are dedicated to language
Bayesian network used for inference and prediction may also attract attention in the future as well because they open the road to reasoning (which is not the case with Deep neural nets).
Be aware of what is going on in the "whole field" is very hard for many reasons:
- An exponential number of papers being published;
- The "quality" of some(many?) papers is questionnable, i.e. impossiblitity to reproduce results. According to the community itself, it is due to the lack of senior reviewers, inflation of ai-related publications/conferences....
The best option is to follow some trustable github pages, authors, labs working in your field of expertise. If there is new breakthrough and they bring new concepts/architectures you will be informed.
Hope that help
Fabrice
Fabrice Noreils
Thanks for info.
Yes actually that is the issue, too many papers.
So you advise to keep on with GitHub, that is what I should do, I guess.
- Badges
- Science topic