Efficient Fpga Implementation Of A Generic Function Approximator And Its Application To Neural Net Computation