The Machine Learning Tribes
The Master Algorithm by Pedro Domingos
In his book ‘The Master Algorithm’, Pedro Domingos dissects the ML trains of thought into 5 tribes/groups.
Symbolists believe in inductive logic, using data and rules to model intelligent systems. This can be computationally intensive at times, subjected to over-fitting and the ‘Bias-Variance’ paradigm. However, experimentation is the key in most of the methods employed here, for e.g. decision trees.
Connectionists are interested in learning how the brain works and mimic the same using neural networks. Deep learning and mapping the brain (Neurosciences) is endogenous to this tribe.
Evolutionaries go even a step further to adopt from nature by emulating the ‘survival of the fittest’ archetype. Natural process by which our genetic material mixes and evolves is bradytelic. But, with the help of Genetic Programming we can expedite the process and take a crack at Machine Learning.
Bayesians rely on the rule of Thomas Bayes from the 18th century. If we observe that for the sun rises in the east for that last n days, our probabilistic inference should be that it would do so again the next day. Further improvements in constructing logic networks (Bayesian Networks) to explain intelligence looks to be very promising for the Bayesians.
Analogizers tend to generate inferences from the samples that can be generalized across domains. Methods like k-Nearest Neighbors and Recommender Systems belong to this class.
Lastly, Pedro’s Alchemy, eponymous to the author, tries to merge all the schools of thought into one, namely that solves the quest for AI — The Master Algorithm. Although, no such things exists at present, speculating about The After-AI World is very interesting indeed.
(PS: this is a personal interpretation of the facts presented by the author)