A note on machine learning

Machine learning," in the abstract, is the ability of computers to improve their performance of a given task, given feedback.

In the practical, machine learning is a set of techniques developed to help computers perform tasks that computers usually find difficult to do, for example, to recognise faces. Not all machine learning techniques actually involve any learning at all, for example, decision trees. Machine learning varies a lot depending on the type of task it is given. A classification task is a task where a label is given to a certain type of data. For example, "Is this picture a giraffe?" is a classification task with two labels -- "giraffe", and "not a giraffe". Classification tasks get more difficult when there are a lot of categories, e.g. "What is this meme?" is a classification task with a lot of categories. Reinforcement learning is a task like "what is the best way to jump over this fence?" Feedback is not immediately calculable, but still exists.

Neural Networks

"Neural networks" are the sexiest, sauciest parts of machine learning.

Neural networks are very popular in the machine learning world, especially for image recognition...