WHITELEY-J@OSU-20.IRCC.OHIO-STATE.EDU (J. Whiteley) (08/07/90)
Can anyone provide pointers to a quantitative test which can be used to determine whether a set of high dimension patterns (eg, d > 20) is linearly separable? I have located a test by Singleton (1962) which can be used on binary data but I need a test for analog (real-valued) data instead. Duda's book (Pattern Classification and Scene Analysis, 1973) suggests that the simplex linear programming method can be used (pp 166-169). However, for a two class problem where the discriminant function is either > or < zero, I fail to see how LP can be used to arrive at a meaningful solution. Any suggestions or help would be greatly appreciated!!! --Rob Whiteley (email: whiteley-j@osu-20.ircc.ohio-state.edu) Dept of Chemical Engineering Ohio State University -------