A model is a representation of some aspect of the real world. Good models are parsimonious, i.e. they focus on the parts that are relevant to them. In „ML, models also represent an aspect of the real world. More specifically, they usually attempt to determine or predict a certain target variable by using input variables. An ML model consists of mathematical functions (i.e. sets of rules) that are supposed to reflect underlying patterns between the target variable and the input variables.
ML models are created through what is called model training. There are a number of model classes in ML and they differ in the way they attempt to depict the real world. Depending on the underlying problem, some model classes are more appropriate to reflect a given process than others.
ML models lie at the heart of the vast majority of today’s AI systems.
ML models are created through a process called model training. During model training, we provide an ML model with historical data from which it can identify and unveil the relevant patterns between independent and output variables. In our height–weight example, the model is the result of training a model on a dataset of a few dozen people. When we train a model, we have to choose a model class. There are different types of model classes in ML, for example, linear models, tree-based models and neural networks. These model classes differ in the way they depict real life.
Consequently, some model classes are especially apt for certain problems. For example, neural networks try to emulate the structure of the human brain to reflect real-world processes and are extremely good at computer vision and natural language problems.