Pairwise Markov chain
Model Families AIMech:PairwiseMarkovChain DRAFT
Aliases: PMC
A hidden-state sequence-model family in which the joint process of a hidden variable and its corresponding observation is first-order Markov, without requiring either marginal process to be Markov or observations to be conditionally independent given the hidden path.
Identity Notes & Concept Scope
Pairwise refers to the hidden–observed pair Z_t=(X_t,Y_t) at each position, not arbitrary pairwise potentials or a graph with two-node cliques. Classical memoryless-emission HMMs are a constrained special case. The concept is distinct from factorial hidden chains, input-conditioned HMMs, longer-lag contextual emissions, mixed-memory predecessor mixtures and pairwise Markov random fields on non-chain graphs. No normalized identity match was found across all 442 pinned AIO labels and all scoped synonyms and 388 preceding native IDs, names and aliases for the three selected concepts. Full-checkout fixed-string searches included ignored and hidden files in both worktree and main, yielding 8 and 8 matched lines; all 8 unique contents were read. Hits are prior pending leads and primary-reference context, not records of these identities. All 78 semantic label/synonym hits and 22 fully annotated exact AIO CLASS nodes with outgoing edges were read. Twenty-six contextual YAMLs were byte-verified after prior full reading and three freshly read, totaling 22 generated and seven native contextual records. Short acronym identities were compared across the complete normalized record inventory. Absence is bounded to the pinned source, maintained collection and inspected semantic neighborhood.
Claims & Grounded Evidence
1 Claim
Mechanism
#method
Specify a normalized initial joint distribution p(z_1) and a normalized joint transition kernel K_t(z_(t+1)|z_t) for z_t=(x_t,y_t). The finite joint law is p(z_1)*product_t K_t(z_(t+1)|z_t). A classical HMM results when the initial law factors as pi(x_1)*b_(x_1)(y_1) and K_t factors as a(x_(t+1)|x_t)*b_(x_(t+1))(y_(t+1)). General pairwise kernels may retain dependencies removed by those restrictions. Given the entire observed sequence y, the posterior over x has a chain factorization even if its unconditioned marginal process is not Markov. Define alpha_1(x_1)=p(x_1,y_1), propagate alpha_(t+1)(x_(t+1)) by summing alpha_t(x_t)*K_t(x_(t+1),y_(t+1)|x_t,y_t), and propagate beta backward with beta_T=1 using the same observed kernel. Normalizing alpha_t*beta_t yields the state posterior; including the intervening kernel yields adjacent-state posterior probabilities. The source also writes the chain density using consistent adjacent joint marginals divided by interior single-position marginals, and discusses iterative conditional estimation from hidden-state posterior expectations or samples.
⚠️ Curated Limitations & Negative Results
- A Markov joint process need not have Markov hidden or observed marginals. Conversely, a special parameter choice can make those marginals Markov; non-Markov behavior is permitted, not required by the definition.
- The source’s adjacent-marginal density ratios require mutually consistent marginals and valid support. Arbitrary pairwise tables cannot be inserted as if they were consistent marginals. A normalized initial law and transition kernels provide an explicit finite-horizon convention.
- The converse characterization in Proposition2.3 uses an additional symmetry hypothesis. It is not an unrestricted criterion for every hidden marginal to be Markov.
- The forward recurrence is used only for t=1,...,T-1; the printed index range beside Equation(3.3) does not correctly delimit that update. Posterior normalization requires positive observation evidence.
- Iterative conditional estimation is not automatically identical to EM and does not, by its name alone, establish likelihood monotonicity, identifiability or global convergence. Stationary parameterizations must preserve the consistency of their pair marginals.
- The primary experiments use specified synthetic chains and a noisy image whose scan is explicitly neither an exact HMC nor PMC. They do not show universal segmentation superiority, and no empirical result or author-code reproduction is curated.
Artificial Intelligence Ontology (AIO) Grounding
Mapped AIO Classes
| AIO Target Class | Mapping Relation | Mapping Rationale |
|---|---|---|
AIO:MarkovChain
Markov Chain
|
broad_match | The defining stochastic state is the complete pair Z_t=(X_t,Y_t), whose next-state law depends only on Z_t. The AIO Markov-chain family is broader than this hidden–observed joint-state construction; the mapping says nothing equivalent about either marginal. |
AIO:BayesianNetwork
Bayesian Network
|
broad_match | For a finite horizon the normalized joint-state chain has an acyclic directed factorization. Splitting each joint kernel by the probability chain rule also gives an ordered within-slice factorization without imposing causal semantics. |
Evidence Sources & Primary Literature
1 Source| Source ID | Publication / Document | Type / Access | Reference & Checksum |
|---|---|---|---|
method-source
|
Wojciech Pieczynski
|
primary paper
full_text
|
ID: doi:10.1109/TPAMI.2003.1195998
sha256:cbe578538a26…
|
Curation Provenance
Curator: Codex
Method: ai_assisted
Model: GPT-6
Created: 2026-10-11T20:49:12Z
Updated: 2026-10-11T20:49:12Z
Source File:
data/curated/PairwiseMarkovChain.yaml