There are many different fields in DL and an extensive research in each one of them.
The main 3 topics are computer vision that mainly use CNN architectures, NLP that mainly use RNN architectures (but not only, recent papers also combine CNNs in many NLP tasks) and general ML topics that use all the combinations of FC(fully connected layers ), CNN and RNN.
At the moment I'm more into NLP so I can share some new algorithms in the field.
One of the most recent hot topics is this:
Better Language Models and Their Implications - https://blog.openai.com/better-language-models/
Another few popular innovations are these:
Bert - https://github.com/google-research/bert
Attention layer - https://medium.com/syncedreview/a-brief-overview-of-attention-mechanism-13c578ba9129
Of course the are many more papers worth mentioning in NLP field, so it is just a small taste. You should be more specific with you question or motivation to get more accurate replies.
Depends on which domain you are particularly looking for.
To cover the NLP space, the most recent massive breakthrough has been what is so called "THE IMAGENET MOMENT FOR NLP". Basically, its introducing effective transfer learning in language model and leveraging them to perform better.
It is really a big leap towards transfer learning for NLP.
You can read more here: https://thegradient.pub/nlp-imagenet/
https://www.analyticsvidhya.com/blog/2018/12/key-breakthroughs-ai-ml-2018-trends-2019/
This page holds a good summary
There are many new contribution on the hybridisation of machine learning like :
* Landslide susceptibility modeling using Reduced Error Pruning Trees and different ensemble techniques: Hybrid machine learning approaches.
* Forecasting performance comparison of two hybrid machine learning models for cooling load of a large-scale commercial building.
* A hybrid machine-learning and optimization method to solve bi-level problems.
* ReviewModus: Text Classification and Sentiment Prediction of Unstructured Reviews using a Hybrid Combination of Machine Learning and Evaluation Models.
There is no specific answer! it depends on the problem that is being solved !
However, one of the latest algorithms is the Teacher-Student Curriculum Learning (TSCL) "curriculum learning" , where in this case, the algorithm is trained to learn on a meaningful order rather than just examples being fed to them.
There are many different fields in DL and an extensive research in each one of them.
The main 3 topics are computer vision that mainly use CNN architectures, NLP that mainly use RNN architectures (but not only, recent papers also combine CNNs in many NLP tasks) and general ML topics that use all the combinations of FC(fully connected layers ), CNN and RNN.
At the moment I'm more into NLP so I can share some new algorithms in the field.
One of the most recent hot topics is this:
Better Language Models and Their Implications - https://blog.openai.com/better-language-models/
Another few popular innovations are these:
Bert - https://github.com/google-research/bert
Attention layer - https://medium.com/syncedreview/a-brief-overview-of-attention-mechanism-13c578ba9129
Of course the are many more papers worth mentioning in NLP field, so it is just a small taste. You should be more specific with you question or motivation to get more accurate replies.
We believe in my mind:the following methods remains a good ways and used in accordance with the type and nature of the data:
Decision tree, Logistic regression
Support vector machine, Naive Bayes
, k-NearestNeighbor, k-means, Markov
Take a look at Capsule Net which was proposed by Geoffrey Hinton.
Article Dynamic Routing Between Capsules
In fact it is a whole new way of looking at DNNs.
Probabilistic Machine Learning:
https://www.tensorflow.org/probability
https://www.youtube.com/watch?v=BrwKURU-wpk
Transformers: https://towardsdatascience.com/transformers-141e32e69591
- Badges
- Science topic