Giving computers the ability to learn from data

3 different types of machine learning

• Supervised learning

  • Learn a model from labeled training data. When the training data is in the form of discrete classes, this task is also called as classification.
  • If the value of a continuous variable is predicted, then it is called as regression which is also considered a type of supervised learning. Here the learning is to create a model that represents a straight line which minimizes average squared distance between the sample points and the fitted line.

• Unsupervised learning

  • Here we are dealing with unstructured and data of unknown structure. We are still able to group data and observe patterns without having to define a training set or reward function.
  • Clustering is a popular example of unsupervised learning where values are grouped into clusters such that they are more similar to their group means than they are dissimilar to members of other groups.
  • Dimensionality reduction is another type of unsupervised learning where data is reduced or compressed to reduce noise and also visualize high order data in 2D or 3D plots.

• Reinforcement learning

  • A system that improves performance or learns through a series of interactions with the environment. After every interaciton, the environment rewards back based on how well the system performed. In this aspect (of feedback), reinforcement learning can be considered as a subset of supervised learning.