The Discipline of Machine LearningΒΆ

Article on Machine Learning.[0]

Last 50 years the field of Machine Learning has gone from the field of games to using statistics in understanding the learning process.

The defining question of Machine Learning is, how can we build computers that automatically improve with experience, what are the fundamental laws that govern all of the learning process?

Given a task T, a performance metric P, and experience E, how does the performance metric P improve with respect to experience E, when accomplishing a task T.

Machine Learning is a natural outgrowth of Computer Science and Statistics. It is both Computer Science and Statistics meeting to exploit the best each has to offer individually.

Machine Learning is also helping us understand animal learning. Specially, Temporal Difference Learning [1], is the concept that is suggested as how humans learn.

Some applications of Machine Learning

  • It was only after 1985 did applications in Machine Learning start coming up.

  • Speech recognition systems is widely used Machine Learning application.

  • Computer Vision is another widely used. USPTO employs a large scale machine learning system to identify addresses from Handwritten notes and sort the envelops.

  • Medical Diagnosis and disease spread based on patient admission and out of counter purchases.

  • MRI scans and understanding of it used ML too.

  • In computer science*

  • ML in computer science is applied for software where traditional algorithms are hard to develop. Like sensor based systems.

  • Robot navigation uses ML.

  • Personalization software uses ML.

  • Constant Learning*

The ongoing research is about creating ML systems that can be like constant self-supervising learners that improve with experience just like humans do.

  • Ethical questions*

In learning with private data, there are ethical questions. We are dealing with it.

[0]: http://www.cs.cmu.edu/~tom/pubs/MachineLearning.pdf

[1]: https://www.wikiwand.com/en/Temporal_difference_learning