Machine learning is the field of artificial intelligence that aim to provide computers with the ability to learn without being explicitly programmed. There are three categories of machine learning methods :
- Supervised learning : The machine is trained on a subset of the dataset containing example inputs associated to their desired outputs. The goal is to find a correlation that maps inputs to outputs. The more the dataset is structured and labelled, the best is the model quality.
- Unsupervised learning : No labels are given to the machine, leaving it on its own to find structure in its input. The goal is to find some hidden patterns or correlations within the dataset.
- Reinforcement learning : The machine evolves in a dynamic environment in which it must solve a specific problem (driving a car, playing a game). The machine gets rewards and punishments while processing the data read from the environment interactions.
Basically, the steps for building a model in machine learning are :
- (1) Analysing the data and chosing the learning strategy
- (2) Preparing the data
- (3) Applying the learning strategy on the training set
- (4) Evaluating the model prediction accuracy on the test set
- (5) Improvement Loop : Correcting/adjusting the parameters of the learning algorithm in step (2) in order to improve the results in step (3) until the model is working properly.