A General Machine Learning Process

From Deep Learning Course Wiki
Jump to: navigation, search

A general machine learning process is built of three core variables:

  • T = Task
  • P = Performance Measure
  • E = Experience

Putting this all together with some guy's super formal definition of Machine Learning, we get:

“A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.” - (Tom M. Mitchell, 1997)

As a practitioner of Machine Learning, you'll need to figure out what these 3 variables are for every project. Here are some examples:

  • The Task, T
    • Determining whether a cucumber is of high quality, given an image
    • Reading the hand written address off of a letter
    • Estimating the price of a house given its amount of square feet
    • Analyzing sentence structure and determining where the subject, verb, and noun are
  • The Performance Measure, P
    • Mean Squared Error
    • Log Loss
    • Cross Entropy
    • Hinge Loss
  • The Experience, E
    • Collection of images
    • Passenger information for an airplane
    • Text from a book
    • Audio from a phone call