Chapter 26: Finite-Blocklength Information Theory

Learning Objectives

• Define channel dispersion and state the normal approximation for the maximal coding rate
• Derive the random coding union (RCU) bound as a non-asymptotic achievability tool
• Derive the meta-converse (hypothesis-testing converse) as a tight non-asymptotic upper bound
• Compute the rate-reliability-blocklength tradeoff for AWGN and binary symmetric channels
• Understand why the classical capacity formula is a poor predictor for short blocklengths ($n \sim 100$-$1000$)
• Apply finite-blocklength analysis to URLLC design and multi-user channels