In the past 32-bit single precision representation for a floating point number was common. In this representation, each value is stored in 4 bytes with 1 bit for the sign, 23 bits for the normalized mantissa, and 8 bits for the BIASED exponent. Given that the exponent cannot contain either all 1's or all zeros, determine a logical choice for the bias. Determine: (a) the machine epsilon, (b) the smallest positive representation, and (c) the largest positive representation