Machine Learning refers to a lot of things in a vast field which is rapidly growing. Arthur Samuel (1901-1990), an American pioneer in the field of computer gaming and artificial intelligence, coined the term “machine learning” in 1959. He defined it as a “field of study that gives computers the ability to learn without being explicitly programmed”. Today, many say that machine learning is the future, although I’d say is already part of lots of things we have today.
Machine learning, as a type of artificial intelligence (AI), enables computers to learn without being explicitly programmed, and to improve their functions when exposed to new data. By analyzing patterns in this data, the machine learning algorithms are self-adjusting based on a set of design rules.
Machine learning algorithms could be grouped based on their:
- Learning style
Algorithms by Learning Style
Two types of machine learning algorithms are commonly used today –supervised and unsupervised. In a supervised learning mode, what has been learned in the past is used to analyze new data, while unsupervised algorithms are capable of inferring from new datasets.
In other words, in a supervised learning algorithm, the machine is trained on known input-out pairs to get the desired output; an inferred function is generated as a result of the training, and could be used on new datasets.
The unsupervised learning algorithms are fed input data without labeled responses, and are capable of extrapolations from their input; cluster analysis is the most common unsupervised learning technique and allows exploratory data analysis for patterns detection.
The semi-supervised learning algorithm uses a mixture of labeled and unlabeled datasets, and learns the structure of the data as well as makes predictions. Regressions and classifications are typical problems solved through this type of machine learning algorithms.
The reinforcement learning method is inspired by behaviorist psychology and from nature. It is used in many disciplines such as game theory, simulation-optimization, statistics, etc. Even your parents used it on you when you were a kid (hopefully only the positive reinforcement type ☺). Reinforcement learning algorithms do not need correct input-output datasets, or correction of suboptimal actions, they are focused on balancing exploration of unknown datasets with the use of current knowledge.
Algorithms by Similarity
Grouping machine learning algorithms by how they work is very useful, but of course some algorithms fit into multiple categories. Here are some examples of similarity-based types of algorithms:
Future articles will explore these similarity-based groups of machine learning algorithms in more detail.
Popular Machine Learning applications:
- Some of the popular fields of applicability for Machine Learning are:
- Text recognition
- Spam filtering
- Computer vision – image and object recognition, photo search/tagging
- Brain-machine interfaces
- Data analytics – extract knowledge from large volumes of data
And here are some very interesting use cases of Machine Learning in our everyday life:
- Natural Language Processing (NLP)
- Voice Recognition and Text-to-Voice
- Self-driving Cars
- Online Search
- Recommendation Systems (ex. Netflix)
- Fraud Detection
- Network and Data Security