【pre】
When I read a public account tweet , there was such a sentence:
Hey, look at this meaning, CV, NLP, RL, GNN are the vertical fields of DL? The other three are familiar, but what is RL?
So I read 1 and 2 and sorted out what I found useful .
【content】
1. The relationship between AI, ML, RL, and DL
(1) AI: Artificial Intelligence (AI) is a new technical science that develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence.
(2) ML: Machine learning (Machine Learning, ML) is trained through algorithms and using historical data, and a model will be generated after the training is completed . When presented with new data, the model produced by training will be used to make predictions.
(3) RL: Representation Learning (RL) is a collection of techniques for learning a feature—it converts raw data into a form that can be effectively exploited by machine learning . It avoids the trouble of manually extracting features, and allows the computer to learn how to extract features while learning to use features: learn how to learn. That is, without manually constructing features .
(4) DL: Deep learning (Deep Learning, DL) is the best embodiment of learning RL - deep learning is used to solve the problem of troublesome and inaccurate manual construction of features. The advantage is that the machine can build features by itself . It is unclear which factor has the greatest impact on the results .
Well, a picture is worth a thousand words.
2. The relationship between CV, NLP, ASR and ES
Let's look at the sentence in the tweet again:
From what he said, it seems that DL has CV, NLP... these fields.
But after understanding, I think it should be more like this :
Computer Vision (CV)
Automatic Speech Recognition (ASR)
Natural Language Processing (NLP)
Expert System (Expert System, ES)
That is, CV, ASR, NLP, ES, etc. are related research in the field of AI, and DL is a method for researching these fields .
【more】
Digression: I got to know a conference, ICLR.
Here are the buzzwords in ICLR 2021:
There is another interesting point of view, to some extent, this is also a trick :).
After reading it, I plan to learn what Reinforcement Learning is, and the GNN that I glanced at before...