02-18-2013 09:24 PM
This is a useful article to read for even for novices at machine learning algorithms.
Domingos takes a practical approach to creating machine learning applications by summarizing 12 key lessons. He outlines three components to learning algorithms: Representation, Evaluation and Optimization giving a helpful table of algorithmic techniques corresponding to each component. Although some of his statements seem like plain old common sense, such "Theoretical Guarantees Are Not What They Seem," Domingos presents a lot of data to back it up. Like any good researcher, he also provides an extensive list of sources. Like any good teacher, he also includes places where a student can learn more.
02-19-2013 11:55 AM
The article raises an interesting (for me) question concerning the evaluation phase. If one uses Content Categorization Studio to identify things to look for in trying to classify something, is there a way to use that tool to exclude things that would provide a misleading direction?