Beyond affinity propagation: message passing algorithms for clustering
Bibliographic record
Abstract
Affinity propagation is an exemplar-based clustering method that takes as input similarities between data points. It outputs a set of data points that best represent the data (exemplars), and assignments of each non-exemplar point to its most appropriate exemplar, thereby partitioning the data set into clusters. The objective of affinity propagation is to maximize the sum of similarities between the data points and their exemplars. In this thesis, we develop several extensions of affinity propagation. The extensions provide clustering tools that go beyond the capabilities of the basic affinity propagation algorithm, and generalize it to various problems of interest in machine learning. We also investigate alternative approaches to the underlying mechanism of affinity propagation using recent inference techniques that are based on optimization theory. Affinity propagation was first described using a particular graphical model for the exemplar-based clustering problem. We first provide an alternative graphical model and derivation of affinity propagation, which are more amenable to model manipulation. Building on this representation, we develop capacitated affinity propagation, semi-supervised affinity propagation, and the hierarchical affinity propagation algorithms. We also discuss the relationship of affinity propagation to some canonical problems in combinatorial optimization. The underlying mechanism of affinity propagation is an approximate inference procedure known as max-product belief propagation. We provide a comparison of affinity propagation to alternative inference techniques such as max-product linear programming, and dual decomposition. We show that for a collection of benchmark data sets, affinity propagation outperforms these more theoretically justified approaches. We conclude by discussing the contributions and findings of this thesis, and how they relate to current research themes in more general inference problems. We point to several interesting avenues for future research.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.005 | 0.017 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.003 | 0.002 |
| Bibliometrics | 0.003 | 0.006 |
| Science and technology studies | 0.002 | 0.003 |
| Scholarly communication | 0.005 | 0.009 |
| Open science | 0.004 | 0.003 |
| Research integrity | 0.004 | 0.006 |
| Insufficient payload (model declined to judge) | 0.005 | 0.003 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".