Classification as an ability is often overlooked, as it is innate in our everyday experiences and it is difficult to identify a specific time when we have not used it. We start grouping objects, colors, letters, and numbers as infants, and this ability becomes more sophisticated, and discerning, as we age. The impact of classification as species is tremendous, even when we don’t realize we are using it at all.
Statistical and computational classification has gone through its improvements as well. Traditional statistics has given rise to Machine Learning, and not so recently, Artificial Neural Networks. These improvements have impacted culture and society tremendously in the last two decades, and this trend will clearly increase in the coming decades. Not all is sunny, as with any new or immature technology there are missteps and mistakes along the way. With greater understanding of how these algorithms and technologies function, the hope is that we will have more of a handle and control of our futures.
In the following paper I describe some of the characteristics, advantages, and drawbacks of popular classification algorithms, and compare their performance across two experiments. By no means are these implementation and observations definitive, the intent of these results is to identify where some methods are more applicable than others, resulting in greater efficiencies when approaching a classification problem.
Paper: Comparison of Classifier Methods
Supporting Notebooks: Image Classification, Text Classification
Vanished Gradient 