Module isse 

ISSE — Indirect Secondary Symbol Estimation model.

Provides model #19 for the mixer: a 3-level ISSE chain using CROSS-CONTEXT hashes that combine dimensions the main n-gram models don’t capture.

The main order-0..9 models use pure n-gram context (c1, c2, …, c9). This ISSE model uses: Level 0 (ICM): word-position context (position within current word + bpos) Level 1 (ISSE): byte-class bigram transition context Level 2 (ISSE): sparse skip-2 context (c1, c3 — skipping c2)

These contexts are orthogonal to the main models, giving the mixer genuinely new information rather than duplicating existing signals.

Architecture from ZPAQ (Matt Mahoney): ICM: context_hash → bit_history_state → probability (StateMap) ISSE: (bit_history_state, p_in) → p_out via learned weights (w0, w1)

Structs

IsseChain

ISSE model: 3-level chain with cross-context hashes.