5. bfloat16 is a half precision floating-point format that is only 16 bits wide, which is widely used by Al processors. The leftmost bit is the sign bit (S), the exponent is 8 bits wide and has a bias of 127, and the fraction is 7 bits wide. Similar to IEEE single precision and double precision, a hidden 1 is assumed. The value is calculated as (-1) (1+fraction) x 2 (exponent-bias). An example of bfloat 16 number is given below. sign exponent (8 bits) fraction (7 bits) 0011111000100000 76 (bit index) 15 14 (a) Write down the 16-bit bfloat16 pattern of decimal number -0.65625. < (b) The exponents of 00000000 and 11111111 are reserved and cannot be used to represent legal numbers. What are the smallest positive number and the largest positive number of bfloat16 format. respectively? Write down their 16-bit patterns.