Every time I hear people saying "AI and Machine Learning" or "Artificial Neural Network and Deep Learning" etc, it sounds to me as "people and men and Donald Trump" as if they are different.
These concepts have containment relationship, as such machine learning is a kind of AI; ANN is a kind of machine learning and deep learning is a kind of ANN. Anyone disagrees?
Artificial intelligence is a branch of computer science that aims to create "intelligent machines".
Unfortunately the founders of AI (John Mc Carthy, Marvin Minsky, Allen Newell and Noam Chomsky) brought the term "Intelligence" but philosphers are still debating about what intelligence means...
The core problems of artificial intelligence includes
_ Knowledge
- Reasoning
- Problem solving
- Perception
- Learning
- Planning
- Ability to manipulate and move object
As you can see machine learning is a branch of "AI". For me the best definition of what Machine learning is, come from Dr Yoshua Bengio: it can be defined as seeking to provide knowledge to computers through data, observations and interacting with the world. That acquired knowledge allows computers to correctly generalize to new settings.
Machine learning is usually composed of three branches:
- Supervised learning
- Unsupervised learning
- Reinforcement learning
Artificial neural networks (ANN) are in the landscape for a while. The earliest ANN can be traced back to 1943 when Warren McCulloch and Walter Pitts developed a computational model for NNs based on mathematics and algorithms called threshold logic and then came the Perceptron in 1958 with Franck Rosenblatt.
But the traditional ANN method suffered from problems such as overfitting, diminishing gradients and thus just few researchers were working on them in the 80s, 90s (Le Cun, Jeoffrey Hinton or Joshua Bengio who are the fathers of Deep learning).
Deep learning uses larger numbers of hidden layers whereas traditional ANNs normally can only afford one or two hidden layers . With the appearence of GPU, tons of data collected from internet and new algorithms (back prop improvement, drop out, polling layer, ReLu...) it was possible to address the overfitting and vanishing gradient issues and you can see now Deep Learning applied in almost everywhere.
If we go back to the different branches of ML, Deep Learning is mainly used in supervised learning and Reinforcement learning.
Hope that help
Fabrice
Artificial intelligence is a branch of computer science that aims to create "intelligent machines".
Unfortunately the founders of AI (John Mc Carthy, Marvin Minsky, Allen Newell and Noam Chomsky) brought the term "Intelligence" but philosphers are still debating about what intelligence means...
The core problems of artificial intelligence includes
_ Knowledge
- Reasoning
- Problem solving
- Perception
- Learning
- Planning
- Ability to manipulate and move object
As you can see machine learning is a branch of "AI". For me the best definition of what Machine learning is, come from Dr Yoshua Bengio: it can be defined as seeking to provide knowledge to computers through data, observations and interacting with the world. That acquired knowledge allows computers to correctly generalize to new settings.
Machine learning is usually composed of three branches:
- Supervised learning
- Unsupervised learning
- Reinforcement learning
Artificial neural networks (ANN) are in the landscape for a while. The earliest ANN can be traced back to 1943 when Warren McCulloch and Walter Pitts developed a computational model for NNs based on mathematics and algorithms called threshold logic and then came the Perceptron in 1958 with Franck Rosenblatt.
But the traditional ANN method suffered from problems such as overfitting, diminishing gradients and thus just few researchers were working on them in the 80s, 90s (Le Cun, Jeoffrey Hinton or Joshua Bengio who are the fathers of Deep learning).
Deep learning uses larger numbers of hidden layers whereas traditional ANNs normally can only afford one or two hidden layers . With the appearence of GPU, tons of data collected from internet and new algorithms (back prop improvement, drop out, polling layer, ReLu...) it was possible to address the overfitting and vanishing gradient issues and you can see now Deep Learning applied in almost everywhere.
If we go back to the different branches of ML, Deep Learning is mainly used in supervised learning and Reinforcement learning.
Hope that help
Fabrice
Deep learning is a subset of machine learning that can solve the problem By using neural networks.
Deep learning is a branch of machine learning algorithms inspired by the structure and function of the brain called artificial neural networks. Some of the deep learning architectures are Deep neural networks, deep belief networks, recurrent neural networks and convolutional neural networks
You are totally right. More here https://www.researchgate.net/post/can_any_clear_the_difference_between_Artificial_neural_network_and_deep_learning#view=5aa90f23615e27e3d8650d08
You are absolutely, right. There are different types of method applied in deep learning of which ANN is a type. so long as the machine is learning from a data and as well it can be technically termed as Machine Learning.
Put it this way. The first weapon we use in Artificial intelligence mission call Machine learning. Machine learning is a fix mathematic formula with programming IF statement to make predictions. Famous algorithm such Naive Bayes, KNN. As time past, we found new weapon name Deep Learning. Deep learning is pure artificial intelligence because it self teaching. It does not rely on fix mathematic formula and programming IF statement to predict. It can contain more than 100 layers to teach itself. The deeper the layer, it can learn better and accurate to make prediction. The popular layer such as Resnet50, InceptionV4, DenseNet.
I think the most documented answer is that of Mr. Fabrice Noreils.
I think, however, that some clarifications are necessary.
I don't think the 'fathers' of AI were wrong. Put it bluntly: Artificial Intelligence means intelligence (and we implicitly think at human intelligence even if is not necessarily true!) realized not by the nature but 'artificial'.
The fact that we still don't have a clear understanding of what intelligence means is another problem.
My understanding of intelligence is that it is the ability to cope with unknown situations. If we think a little in detail we see that this 'definition' encompasses all mentioned 'core problems' except the last one (Ability to manipulate and move object) which, of course, is useful but is a physical ability.
As regards Machine Learning is certainly part of AI: is the artificial learning. And, by the way, those three 'branches' are, in fact, methods of human learning which we strive to transfer to artificial intelligence.
ANN and Deep Learning are techniques not abilities, features of AI.
The literature was analyzed, which contains all the names listed at once by Sisay Chala. These are articles found in the Web of Science on request
ts=(Artificial intelligence Machine Learning Artificial Neural Network Deep Learning) - they are 427 articles.
file topic_report.docx = 20 topics from 427 articles which have words
ts=(Artificial intelligence Machine Learning Artificial Neural Network Deep Learning)
in their titles, abstracts or keywords. Each topic is represented through 20 words and 20 thrases with which it is discussed in these articles. Really this terms are the names of methods, objects, properties, laws and so on for topic in question. In addition each topic has quotes from two articles in which it is most manifested.
file a1_basic.xlsm - articles with basic knowledge on the topic
file a3_novelty.xlsm - articles potentially with novelty
We conducted an investigation in which we examined all terms that might have something to do with AI. We included many terms, for example, business intelligence, where most would claim that has not much to do with AI or neural networks. We are currently working on a paper in which we will present the results, shortly. In fact, it seems that as time goes by, all terms become more and more intertwined through intersecting themes, similar to a symbiogenesis in nature. First preliminary findings can also be found in our paper at IEEE https://ieeexplore.ieee.org/document/8718286
A real taxonomy or ontology is surely needed. As with ordinary dictionary definitions there are synonyms and antonyms. A study of typical usage would also be helpful in this regard.
- Badges
- Science topic