Unweighted Multiview Convex Mixture Clustering
Algorithms AIMech:UnweightedMultiviewConvexMixtureClustering DRAFT
Aliases: Multiview Convex Mixture Model Clustering, Convex Mixture Models for Multi-view Clustering
A multiview exemplar-clustering procedure that fits a single vector of exemplar weights by minimizing the equally weighted sum of fixed-component negative log likelihoods across aligned data views, then selects a specified number of exemplars.
Identity Notes & Concept Scope
This is the unweighted Tzortzis-Likas 2009 procedure. Shared q couples corresponding exemplar indices across views, while each view retains its own fixed component family and scale. It averages view-specific log objectives and responsibilities; fixing uniform view weights inside the later weighted log-of-sum objective does not recover this procedure. The original sum-of-posteriors hard assignment and top-M ranking stop are retained explicitly. No normalized identity collision across all 442 pinned AIO IDs, labels/scoped synonyms and 487 maintained IDs, names and aliases. Full-checkout searches include ignored and hidden files: 8 matched lines in the worktree and 9 in main. All six distinct matched texts were read; they concern the prior exemplar-clustering frontier, its source/query URLs and an unrelated FreeFem lexer token. Seven annotated AIO nodes and seven full native neighbors were inspected. The same pinned bytes retain the previously read 442 class labels and 138 semantic label/synonym rows. Absence is bounded to this snapshot and corpus.
Claims & Grounded Evidence
3 Claims
Mechanism
#shared-weight-fitting
For N observations with V aligned views, fix view-specific component values f_vij=f_vj(x_vi) and common simplex weights q_j. Minimize -(1/(N V))sum_vi log Q_vi, where Q_vi=sum_j q_j f_vij. Set q_j_new=q_j/(N V)*sum_vi f_vij/Q_vi. Each view contributes its average current responsibilities equally. Common factors depending on the view and observation but not the exemplar cancel in this update. With finite positive component values, positive support and exact unpruned iteration, the soft convex problem attains a global optimum; coefficient uniqueness is not implied.
Mechanism
#specified-exemplars-and-assignment
Given a requested M, retain the M largest fitted q values as exemplars, each anchoring its own cluster. Assign every other observation to the selected exemplar maximizing sum_v q_j f_vij/Q_vi. This sums per-view posteriors, rather than multiplying densities or taking the maximum over individual views. The source uses strict comparisons without a tie policy. Its implementation stops when the selected top-M indices and their ranking stay unchanged for several consecutive iterations; this can occur before the soft objective is optimized.
Limitation
#scope-and-calibration
View families and beta_v values remain supplied choices. The empirical scale N^2 log(N)/sum_ij d_vij is a search guide and is undefined when its denominator vanishes. The fitting update needs O(N^2 V) dense arithmetic per iteration, apart from distance construction. A fixed-M hard partition has no transferred global-optimum guarantee; stable ranking is not a bound on residual objective error. Equal view contribution also does not guarantee robustness to uninformative or systematically biased views.
⚠️ Curated Limitations & Negative Results
- The record uses the finite evaluated-component/log-likelihood objective; the source empirical-versus-continuous KL shorthand requires a compatible measure or lift. Gaussian exponents use squared Euclidean distances.
- The soft fit is a shared-weight specialization of fixed-component EM. The complete multiview data construction, equal-view objective and sum-of-posteriors/top-M policy define this narrower algorithm.
- Source experiments remove observations lacking required views and select scale multipliers by label entropy. They do not establish an intrinsic missing-view algorithm, fully label-free model selection, or universal performance. No historical experiment was reproduced.
Artificial Intelligence Ontology (AIO) Grounding
Mapped AIO Classes
| AIO Target Class | Mapping Relation | Mapping Rationale |
|---|---|---|
AIO:UnsupervisedClustering
Unsupervised Clustering
|
related_match | The procedure fits exemplar weights and constructs a grouping from unlabeled observations. The inspected AIO target is a clustering task; this algorithm-to-task association asserts neither equivalence nor an upstream subclass. |
Evidence Sources & Primary Literature
1 Source| Source ID | Publication / Document | Type / Access | Reference & Checksum |
|---|---|---|---|
multiview-source
|
Grigorios Tzortzis, Aristidis Likas
|
primary paper
full_text
|
ID: doi:10.1007/978-3-642-04277-5_21
sha256:d0bde397e858…
|
Curation Provenance
Curator: Codex
Method: ai_assisted
Model: GPT-6
Created: 2026-10-12T05:43:14Z
Updated: 2026-10-12T05:43:14Z