Definition:
According to Arthur Samuel, machine learning is the field of study that gives computer the ability to learn without being programmed explicitly.
Simply it is just making the computer learn from the data.
History:
Although it is being extensively used after the 2000s, it is said that some of the statistical methods were discovered in early 1950s. Simple algorithms were used in 1950s. Techniques like support vector clustering and unsupervised learning were widespread in 2000s, while deep learning techniques become popular after 2010 due to the discovery of powerful computers.
Example:
A machine learning algorithm which detects spam email.
Why use machine learning?
- Easy to create and update rules.
Consider the example of spam mail detection system and we are not using machine learning algorithm. We programmed to flag all the emails as spam which contains the word ‘4U’. But the spammers found that all of the emails containing the word ‘4U’ are flagged as spam and they started to use ‘For U’ instead of ‘4U’. A traditional spam detection system would need to program for to detect and flag the emails containing the word ‘For U’. But in the case of Machine learning algorithm, it would learn automatically and start to flag the emails containing the word ‘For U’ as spam. - In complex areas where the traditional approach is too complex or no known algorithm available.
Example: Speech recognition system. - Changing conditions: Machine Learning algorithm can adapt to changes and new data.
- Complex problem and large data applications.
Common examples of machine learning:
- Voice detection by Google Assistant
- Photo tagging in Google Photos
- Google Search result
- Credit card fraud detection
- Movie suggestion engine of Netflix
- Frequent brought together product suggestion in Amazon