In this paper, we propose spatial propagation networks for learning the affinity matrix for vision tasks. We show that by constructing a row/column linear propagation model, the spatially varying transformation matrix exactly constitutes an affinity matrix that models dense, global pairwise relationships of an image. Specifically, we develop a three-way connection for the linear propagation mod...

Structured and semi-structured data describing entities, taxonomies and ontologies appears in many domains. There is a huge interest in integrating structured information from multiple sources; however integrating structured data to infer complex common structures is a difficult task because the integration must aggregate similar structures while avoiding structural inconsistencies that may app...

The Affinity Propagation (AP) is a clustering algorithm that does not require pre-set K cluster numbers. We improve the original AP to Map/Reduce Affinity Propagation (MRAP) implemented in Hadoop, a distribute cloud environment. The architecture of MRAP is divided to multiple mappers and one reducer in Hadoop. In the experiments, we compare the clustering result of the proposed MRAP with the K-...

We propose using the affinity propagation (AP) clustering algorithm for detecting multiple disjoint shoals, and we present an extension of AP, denoted by STAP, that can be applied to shoals that fusion and fission across time. STAP incorporates into AP a soft temporal constraint that takes cluster dynamics into account, encouraging partitions obtained at successive time steps to be consistent w...

Affinity propagation (AP) was recently introduced as an unsupervised learning algorithm for exemplar-based clustering. We present a derivation of AP that is much simpler than the original one and is based on a quite different graphical model. The new model allows easy derivations of message updates for extensions and modifications of the standard AP algorithm. We demonstrate this by adjusting t...

AFFINITY PROPAGATION: CLUSTERING DATA BY PASSING MESSAGES Delbert Dueck Doctor of Philosophy Graduate Department of Electrical & Computer Engineering University of Toronto 2009 Clustering data by identifying a subset of representative examples is important for detecting patterns in data and in processing sensory signals. Such “exemplars” can be found by randomly choosing an initial subset of da...

– Partitioning clustering algorithms, construct non-overlapping clusters such that each item is assigned to exactly one cluster. Example: k-means – Agglomerative clustering algorithms construct a hierarchical set of nested clusters, indicating the relatedness between clusters. Example: hierarchical clustering • In classification, we partition data into known labels. For example, we might constr...

Document clustering as an unsupervised approach extensively used to navigate, filter, summarize and manage large collection of document repositories like the World Wide Web (WWW). Recently, Document clustering is the process of segmenting a particular collection of texts into subgroups including content based similar ones. The purpose of document clustering is to meet human interests in informa...

In our research we sought to create implementations of several common clustering algorithms and a relatively new approach called Affinity Propagation. Our objective was to compare the techniques by running tests on one and two dimensional datasets provided by Professor Trono. Dave Kronenberg implemented a standard randomly seeded K-Means clustering program and many related support functions. We...

In spite of over a century of research on cortical circuits, it is still unknown how many classes of cortical neurons exist. In fact, neuronal classification is a difficult problem because it is unclear how to designate a neuronal cell class and what are the best characteristics to define them. Recently, unsupervised classifications using cluster analysis based on morphological, physiological, ...