Theoretical Investigation of the Minimax Probability Machine
(I. King)

Minimax Probability Machines (MPM) have been popularised recently due to its ability to perform excellent classification tasks. Our project proposes to examine variations of the MPM, e.g., the Biased Minimax Probability Machine (BMPM), the Minimum Error Minimax Probability Machine (MEMPM), the Maxi-Min Margin Machine (M4), etc. to find out the limitations and performances under various data inputs and conditions. The work will benefit those who are designing classifiers for domain specific tasks such as web page categorization, face recognition, biomedical diagnosis, etc.