Machine Learning Algorithm Taxonomy
/:
Over the years there have been a proliferation of algorithms used in AI, so it's hard to understand the technology. I recently came across this taxonomy chart from a guy out of Australia, Jason Brownlee, at a company called Machine Learning Mastery. He offers a guided course to familiarize you about machine learning, along with several books on the topic. His taxonomy of Machine Learning algorithms which he calls a machine learning mind-map, provides a pretty concise snapshot of the options:
Setting aside the PhD-level machine learning classes necessary to actually use them, if you are just looking for the high level snapshot, below is the rundown by category.
Keep in mind that the basic operating premise for all machine learning is that all of these techniques start with a (big) set of data that is used to develop or train a model that is in turn used to determine an output from a new small instance of data. For example, use clustering analysis to develop five (5) customer segments and individualized marketing experience for each cluster based on their product interests, past buying history, and demographics. Then determine which segment a new prospect (visiting your website) is in and present the targeted and personalized marketing experience for that segment. The segments are the model, the next new customer is the new instance of the data.
- Regression: A statistical technique used for prediction and forecasting that finds the relationship between input and output variables. For example, we can figure out the relationship between car sales and car loan interest rates. Once we know that relationship we can predict car sales based on a particular level of loan interest rate.
- Rule System: An algorithmic technique for classification that creates one rule for each predictor in the data, and then selects the rule with the lowest error
Regularization
Neural Networks
Ensemble
Deep Learning
Clustering
Instance Based
Dimensionality Reduction
Decision Tree
Bayesian