Questions: Entropy Rate of Stochastic Processes

3 questions to test your understanding

Score: 0 / 3
Question 1 Multiple Choice

English text has an alphabet of 27 characters (26 letters + space). If characters were i.i.d. uniform, H = log2(27) ≈ 4.76 bits/character. Shannon estimated the entropy rate of English at about 1.0-1.5 bits/character. Why the huge gap?

AEnglish uses fewer than 27 characters in practice
BEnglish has massive redundancy: character frequencies are highly non-uniform, and sequential dependencies (digrams, trigrams, word structure, grammar, semantics) reduce the conditional entropy far below the marginal entropy
CShannon's estimate was inaccurate
DThe i.i.d. model is correct for text; the gap is due to measurement error
Question 2 True / False

For a stationary process, both limits lim (1/n)H(X_1,...,X_n) and lim H(X_n | X_{n-1},...,X_1) exist and are equal.

TTrue
FFalse
Question 3 Short Answer

Compute the entropy rate of a binary Markov chain where P(0|0) = 0.9, P(1|0) = 0.1, P(0|1) = 0.5, P(1|1) = 0.5, and explain what the result tells you about compressing sequences from this chain.

Think about your answer, then reveal below.