We have found that several variants of deep learning have been obtained in the beginning years. Now the question is what is the next variant of deep learning is to come out. Of course one can think of some hybrid models, some of which are already in existence like the fuzzy deep neural networks,
I think, what will be done in deep learning is suggesting new architectures such as Birdnet+.
Preprint BirdNet+: End-to-End 3D Object Detection in LiDAR Bird's Eye View
Talking from personal experience, I think the hype associated with rebranded Neural networks (Deep learning) will vanish to some extent. The biggest problem with deep learning models is their explainability and interpretability. This is a new frontier in Deep learning and hopefully,, we will see more findings in this area. Also Gradient boosters (XGboost for e.g) have comparatively attracted less attention. Gradient boosters perform far better than any deep learning model on structured data. and most real data is structured data.
- Badges
- Science topic
- Similar topics
- Cognitive Science
- Thinking