For two (or more) correlated discrete memoryless sources observed separately and encoded independently without inter‑encoder communication, the Slepian–Wolf Theorem characterizes the achievable region of lossless compression rate tuples when joint decoding is allowed. For two sources X and Y the asymptotically achievable region is R_X ≥ H(X|Y), R_Y ≥ H(Y|X), and R_X+R_Y ≥ H(X,Y), where H denotes Shannon entropy; these bounds show the sum rate can approach the joint entropy even though encoders operate independently.

Principle

Statistical dependence between separately encoded sources can be exploited at joint decoding so that the sum of independent encoder rates need not equal the sum of individual entropies; separate encoding with joint decoding can achieve the joint‑entropy sum rate asymptotically.

Demonstration

Illustrative scenario: Two sensors measure correlated binary sequences X^n and Y^n. Each sensor encodes its sequence at rate just above H(X|Y) and H(Y|X) respectively and sends indices to a central decoder. The joint decoder uses the correlation structure to recover both sequences with vanishing error as n→∞, even though encoders did not share their data or coordinate codewords.

Misapplication

Assuming Slepian–Wolf applies without qualification to one‑shot settings, sources with memory, continuous alphabets, or lossy compression: the theorem is asymptotic and specific to lossless coding of discrete memoryless sources with joint decoding. Also mistaken is the belief that no coordination is ever needed in practice—practical schemes often require shared codebooks or common randomness to approach the bounds.

Consequence

Slepian–Wolf establishes that distributed sources can be compressed nearly as efficiently as centralized ones when joint decoding is available, motivating distributed source coding, network compression strategies and practical codes that approach the bounds.

Reversal

If encoders may communicate, sum‑rate can be reduced further by cooperation; if only separate decoding is allowed, rates must meet individual entropies. For lossy distributed compression, different bounds apply (e.g., Wyner–Ziv and other distributed rate–distortion results), and for non‑i.i.d. sources or finite blocks the asymptotic region changes.

Boundary

Clearly within: discrete memoryless (i.i.d.) sources, lossless recovery requirement, asymptotically large blocklengths, and joint decoder access to both encoded indices. Boundary case: ergodic but dependent sources where single‑letter expressions may require care. Clearly outside: one‑shot lossless coding without joint decoding, lossy distributed compression, and continuous alphabets without quantization.

Semantic Tension

Encoder independence versus decoder cooperation: Slepian–Wolf shows that maintaining independent encoders can still achieve centralized efficiency only when the decoder can jointly process indices—practical systems balance the cost of coordinating codebooks against the benefit of lower transmission rates.

Synthesis

Slepian–Wolf demonstrates that correlation is a decoder‑side resource: separate encoders need not forfeit asymptotic compression efficiency if the decoder jointly exploits dependencies, but realizing that efficiency in practice requires mechanisms (shared codebooks, coordination) to approximate the theoretical construction.