Circuit Synthesis of piecewise-linear multivariate functions
The problem of finding small-size and low-power circuit implementations of multivariate nonlinear functions is of major importance in many applications such as nonlinear control systems, sensor networks (to compress information coming from different measures and thus to save energy), circuit emulation of networks of nonlinear dynamical systems (e.g., biological neurons). This problem can be divided into two strictly related subproblems. The first lies in finding approximate models (belonging to a well-defined class) of either known functions or, at least, functions where a reasonable number of (noisy) measure samples are available (regression set). The second is the definition of a circuit synthesis technique that can be applied to a large class of models.