Go Back
Deep Learning

Machine Learning

Machine learning is the automated process of building new models. It’s called “learning” because it involves data streaming into a system which contains existing data, allowing those two sets of data to combine, recognize new patterns, adjust assumptions, and make something new. In other words, data that learns from data.

The term "Machine Learning" was coined in 1959 by Arthur Samuel and is broadly classified into four distinct types of learning.

  • Supervised Learning – building a statistical model from a series of labeled input and output data sets
  • Unsupervised Learning – extracting insights exclusively from input data with no predetermined or desired outcome
  • Semi-supervised Learning – combining the former two methods where only some of the outputs are known
  • Reinforcement Learning – developing models from reward systems in a dynamic environment

Similar to human learning, the goal of machine learning is to generalize from learned experiences. A machine learning model attempts to evaluate a new input and draw conclusions for an intended output. The results can be evaluated automatically, or as part of a human-in-the-loop process, to continuously improve the model over time.

In general, the secret to successful machine learning is to continuously feed data into an algorithm. Many exciting advances in machine learning have come from applying greater quantities of data and processing power to pre-existing algorithms, rather than creating brand new algorithmic approaches.

Resources

ML ClassificationDo we Need Hundreds of Classifiers to Solve Real World Classification Problems?

LINK

TensorFlowOpen-source software library for dataflow and differentiable programming

LINK

Scikit-learnMachine learning library for the Python programming language

LINK