What is machine learning?
The term machine learning is an umbrella term for many facets of learning in the field of AI. It is a subfield of computing and a branch of artificial intelligence specializing in the development of techniques to promote the learning of intelligent agents.
Machine learning is a fundamental part of Big Data. The term was invented in 1959, but it has been gaining more and more relevance in recent years.
What is an intelligent agent?
Although the term intelligent agent can be applied to living organisms, it is used in computing to denote entities capable of perceiving their environment, processing information that is captured, and responding rationally.
When is an intelligent agent said to be learning?
An intelligent agent is considered to be learning when it improves its performance through experience – that is, using skills that were not present in its original programming.
Machine learning spans three types of learning:
- Growth: acquiring information from the environment and storing it in memory.
- Restructuring: interpreting data generates new knowledge.
- Adjustment: generalizing several concepts or creating its own concepts.
What are the main types of machine learning?
- Supervised learning: in this type of learning, the algorithm is provided with a tagged set of input data; that means it is accompanied by the corresponding results that are desired. The learned function should be used to make predictions about new untagged data.
- Unsupervised learning: in unsupervised learning, an untagged data set is provided for the algorithm to discover patterns, hidden structures, or groupings the set contains and organize them.
- Reinforcement learning: in reinforcement learning, the agent interacts with the environment, performs actions, and receives feedback in the form of rewards or penalties. As the agent explores the environment, it learns to make sequential decisions to accumulate rewards and develops a strategy to maximize the reward in the long term.
There are also hybrid models, such as semi-supervised learning, transfer learning, transduction, etc.
What types of algorithms are used in machine learning?
Several types of algorithms are used for machine learning. These are some of the many types:
- Clustering algorithms: these divide a set into groups according to its characteristics; for example, k-means, K-NN.
- Decision trees: these are hierarchical structures that make decisions based on conditional rules. They illustrate the possible outcomes of a decision through tree-shaped flowcharts.
- Neural networks: ANNs (artificial neural networks) are used to model non-linear relationships with high-dimensional data (data that has more characteristics than observations). They draw inspiration from biological systems, such as the brain, and are a set of interconnected layers that work collaboratively.
- Deep learning algorithms: these use several layers of neural network algorithms.
- Genetic algorithms: these are inspired by biological evolution and are used to solve optimization problems.
- Bayesian networks: they model probabilistic relationships for decision-making.
What applications does machine learning have?
The applications of machine learning are virtually endless, but these are some of them:
- Smart cities.
- Social networks.
- User experience.
- Medical diagnosis.
- Natural Language Processors.