What is Machine Learning?

1.-Introduction to Machine Learning

Machine Learning is a powerful subset of artificial intelligence that allows systems to learn from data and make predictions or decisions without explicit programming. In that sense, Machine Learning algorithms are able to generalize from sample data

As a descriptive example, imagine a set of images with labels that can define and/or classify the image. A Machine Learning algorithm could define a new image classifier. This new classifier, once trained, should be able to accurately apply labels to new images that have not been seen before.

These machine learning algorithms learn rules from labeled examples. The set of labeled examples used for learning is called training data. The learned rules should be able to generalize to correctly recognize or predict new examples not found in the training set.

Machine Learning combines statistics, computer science and other sciences:

• Statistical methods
• Provides conclusions from data
• Estimate the reliability of the algorithm's predictions

• Computing
• Large scale architectures
• Algorithms for capturing, manipulating, indexing, combining, retrieving, and making predictions on data

• Economics, biology, psychology
• For example: How can an individual or system efficiently improve its performance in a given environment?

Some examples of the applicability of Machine Learning can be the following:

• Finance: Fraud detection in credit card transactions
• Web search: spell checking, result ranking, content sorting and selection, advertising placement
• Voice recognition
• E-commerce: product recommendations
• Spam email filtering

2.- Key concepts of Machine Learning

Supervised Machine learning: The Machine Learning algorithm learns to predict target values from labeled training data. It can be classified into:

• Classification: When the values obtained by the prediction algorithm are discrete. That is, the results can only be obtained for a certain number of values.
  • Example: Let's imagine classifying 4 types of fruits from training data such as size, color, texture...

Figure 1 - Classification Example

• Regression: When the values obtained by the prediction algorithm are continuous. That is, the results can only have any value.
  • Example: Let's imagine determining the price of a home from training data such as: city, neighborhood, m2 , number of rooms, number of bathrooms, CPI from previous years.

Unsupervised Machine learning: The Machine Learning algorithm learns to predict target values from searches for structures in unlabeled data. It can be classified into:

• Clustering: Finding groups of similar instances in data
• Outlier Detection Searching for unusual patterns

Figure 2 - Examples of unsupervised machine learning

Machine Learning algorithms are able to generalize from existing sample data