0:17:11: Q1. How many length-N typical sequences we have for DMS X? ANS: \(2^{NH(X)}\)
0:25:07: Q2. How many bits to store the coded index? ANS: NH(X)
0:31:44: Q3. What is H(X) in terms of p(x)? ANS: \(E[-log(p(X)]\) OR \(\sum_{x} -p(x) log p(x)\) (Be careful that answer is case-sensitive)
0:49:14: Q4 what is the code word for 1? ANS: 1111
0:56:00: Q5 is the code prefix free? ANS: no
1:02:41: Q5 what will be the encoded output in this case? ANS: 1111010111
1:20:23: Q6. What is the respective interval of codeword 11? ANS: [0.75,1]
1:53:34: Q7 What is the length of the interval if the code length is l? ANS: \(2^{-l}\)
2:07:14: Q8 what will be the upper bound of the resulting rate per symbol if we group N symbols before feeding to SFE code? ANS: \(H(X) + 2/N\)
2:28:50: Q9 What is the most probable sequence? ANS: 1000 Tails
2:36:00: Q10 In general, what would be a lower bound of H(X)? ANS: 0
2:40:51: Q11 Is this other function convex? ANS: no
2:43:01: Q12. Should a) E[f(X)]> f(E(X)) or b) vice versa? ANS: a
2:53:25: Q13, what is the h(X)? ANS: log(a)