Machine learning

Machine learning gives computers the ability to learn without being explicitly programmed (Arthur Samuel, 1959).^[1]^[2] It is a subfield of computer science.^[3]

The idea came from work in artificial intelligence.^[4] Machine learning explores the study and construction of algorithms which can learn and make predictions on data.^[5] Such algorithms follow programmed instructions, but can also make predictions or decisions based on data.^[6]^:2 They build a model from sample inputs.

Machine learning is done where designing and programming explicit algorithms cannot be done. Examples include spam filtering, detection of network intruders or malicious insiders working towards a data breach,^[7] optical character recognition (OCR),^[8] search engines and computer vision.

Using machine learning has risks. Some algorithms create a final model which is a black box.^[9] Models have been criticized for biases in hiring,^[10] criminal justice,^[11] and recognizing faces.^[12]

Overview

Learning algorithms try to predict what will happen in the future with patterns from the past. These predictions can be obvious: for example, if the sun rose for the past 10,000 days, it will probably rise again. These predictions can also be more complex. An example of a complex prediction is facial recognition (knowing who someone is by looking at face).

Machine learning programs can do things that is hasn't been told to do by a programmer. Machine learning programs will be shown some patterns. These patterns will be an input (such as a question) and an output (the answer to the question). Then, the machine learning program will predict the output based on the input. Machine learning isn't always necessary. Computers can do simple tasks by being told instructions. However, sometimes there are a lot of things that control the output. Then, it is hard for a human to tell the computer all of the instructions. It is sometimes easier to tell the computer how to teach itself.^[13]

There are a lot of different ways to tell the computer to teach itself. When a problem has a lot of answers, different answers can be marked as valid. This is used to form data that the computer is trained with. One example of training data is the MNIST data. The MNIST data has images of handwritten numbers. The computer can learn to identify handwritten numbers using the MNIST data.

Machine Learning Media

Machine learning as subfield of AI
A support-vector machine is a supervised learning model that divides the data into regions separated by a linear boundary. Here, the linear boundary divides the black circles from the white.
Clustering via Large Indel Permuted Slopes, CLIPS, turns the alignment image into a learning regression problem. The varied slope (b) estimates between each pair of DNA segments enables to identify segments sharing the same set of indels.
An artificial neural network is an interconnected group of nodes, akin to the vast network of neurons in a brain. Here, each circular node represents an artificial neuron and an arrow represents a connection from the output of one artificial neuron to the input of another.
A decision tree showing survival probability of passengers on the Titanic
Illustration of linear regression on a data set
A simple Bayesian network. Rain influences whether the sprinkler is activated, and both rain and the sprinkler influence whether the grass is wet.
An example of Gaussian Process Regression (prediction) compared with other regression models
The blue line could be an example of overfitting a linear function due to random noise.