Landmark Recognition Using Ensemble-Based Machine Learning Models

Machine Learning And Its Techniques

Figure 1.

An overview of machine learning models

Machine Learning can be defined as the field of Artificial Intelligence (AI) that deals with training machines to determine patterns in the data that is fed to it and based on the patterns that the machine determines, making acceptable decisions on the new data which may be used to further carry out different other tasks. Machine learning (ML) algorithms are mostly thought to work well only on very crisp and clean data. Moreover, the lesser the data needed for generalization, the better is an ML model expected to work (Jordan & Mitchell, 2015).

Machine Learning models are trained on various types of data. Based on the type of data, ML techniques are classified into the following categories:

Figure 2.

Different types of machine learning

• 1.

  Supervised Learning: This type of learning is the “Help would be appreciated” type of learning. In this type of learning, we provide data for training that is already labeled. A dataset that includes both inputs for training and their real outputs is given. Then, the model is trained which can predict the outcomes of new data to be given (Caruana & Niculescu-Mizil, 2006).

• 2.

  Unsupervised Learning: This is the “Help is not needed” type of learning. In this type, we do not provide a labeled data for training. The ML models are trained on datasets without outputs and then other parameters determine how well the model has been trained (Barlow, 1989).

• 3.

  Reinforcement Learning: This is the “I’m learning from the environment” type of learning. In this type of learning, we do not provide data with labeled outputs, but the models are made to learn from external sources. This type of model can train itself over longer periods and get better without worrying about overfitting (Sutton & Barto, 2018).