Is there a reasonably accepted/shared view on how to minimally classify the various fields of AI? There are hundreds of techniques and I have not been able to find a shared exhaustive classification in reasonably disjointed 'fields', or approaches, or 'schools'.
One possible classification (please, feel free to correct my imprecise terminology) is given by Wikipedia in the Artificial Intelligence page
• Symbolic.
• Deep Learning.
• Bayesian Networks.
• Evolutionary.
But it seems to be somewhat lacking (for example, not all NN are deep learning, and where do SVN methods fit here? Bayesian networks seem too specific to accomodate them).
On the "Outline of Artificial Inteligence" page, wikipedians are giving a more detailed view, but it does not look minimalist to me (moreover, 13 Search a field of its own? Is it not 'dispersed' in all other fields? see Note)
A slightly more minimalistic, yet apparently complete, classification can be found in the laymen book "Artificial Intelligence for Dummies". Here the authors mention five "tribes" (I might have placed them in a different order):
• Symbolists: The origin of this tribe is in logic and philosophy. This group relies on inverse deduction to solve problems.
• Bayesians: This tribe's origin is in statistics and relies on probabilistic inference to solve problems.
• Analogizers: The origin of this tribe is in psychology. The group relies on kernel machines to solve problems.
• Connectionists: Thia tribe's origin is in neuroscience and the group relies on backpropagation to solve problems.
• Evolutionaries: The evolutionaries tribe originates in evolutionary biology, relying on genes e programming to solve problems.
Is this the most comprehensive yet mininalist classification? Is there an even more comprehensive one, along these lines? It also seems to me that Bayesian and Analogizer can both fall into the realm of "Statistical Learning".
What I am looking for, in order to orient myself, is some sort of big Venn diagram or better yet partition with all the approaches/tribes (the more general, the better) and examples of techniques (for example, Naive Bayes under Bayesian: Expert Systems under Symbolists: Perceptrons, Neural Networks under Connectivists...)
To be clear, I am not trying to accomodate transversal approaches, such as Classical vs Hachine Learning, or Supervised va Unsupervised vs Reinforcement Learning, or Strong vs Weak AI..
I understand there are hybrid approaches, of course. What I am interested in is the possibly most disjointed, 'pure approaches on the line of the five tribes described above. Am I perhaps looking for a magical unicorn?
Note: As stated above, I am also a bit confused about how to fit searching algorithms, because I tend to see them as either transversal to the various tribes' or even orthogoal to them imaybe part of Optimization goals, along with Predicit ion, Classification and... what else? Planning?), but this might be the subject of another question
