Ill Posed Problem

A problem is ill-posed if it fails to meet one or more of the three conditions of a well-posed problem defined by Hadamard:

  • Existence: A solution exists.
  • Uniqueness: The solution is unique.
  • Stability (continuous dependence): the solution changes continuously with small changes in the input.

Intuition: theoretically solutions exists but tiny disturbances makes them unstable.

Examples:

  • ML1 - Well and Ill Posed Problems “classification is an example of ill posed problem”
  • Computer Vision is another ill posed problem
  • Inverse problems like deblurring an image because small noise can completely change the reconstructed signal.
  • Inverting a covariance matrix of features when the number of features exceeds the number of samples is ill-posed. Since the matrix may be singular or nearly singular making solutions unstable.