Approximate Synthesis for Classification Functions

A classification function is a multi-valued function, where the function values for only a fraction of the input combinations are defined. Many variables in such a function are redundant, and can be eliminated to reduce the circuit size to implement the function. One can further reduce the number of variables by using linear transformation. However, by relaxing the requirement for full accuracy, we can further reduce the number of variables. This paper derives expected number of errors for approximate logic synthesis. Experimental results for various functions are shown.