No abstract

Whether overt attention in natural scenes is guided by object content or by low-level stimulus features has become a matter of intense debate. Experimental evidence seemed to indicate that once object locations in a scene are known, salience models provide little extra explanatory power. This approach has recently been criticized for using inadequate models of early salience; and indeed, state-of-the-art salience models outperform trivial object-based models that assume a uniform distribution of fixations on objects. Here we propose to use object-based models that take a preferred viewing location (PVL) close to the centre of objects into account. In experiment 1, we demonstrate that, when including this comparably subtle modification, object-based models again are at par with state-of-the-art salience models in predicting fixations in natural scenes. One possible interpretation of these results is that objects rather than early salience dominate attentional guidance. In this view, early-salience models predict fixations through the correlation of their features with object locations. To test this hypothesis directly, in two additional experiments we reduced low-level salience in image areas of high object content. For these modified stimuli, the object-based model predicted fixations significantly better than early salience. This finding held in an object-naming task (experiment 2) and a free-viewing task (experiment 3). These results provide further evidence for object-based fixation selection--and by inference object-based attentional guidance--in natural scenes.

For high-dimensional datasets in which clusters are formed by both distance and density structures (DDS), many clustering algorithms fail to identify these clusters correctly. This is demonstrated for 32 clustering algorithms using a suite of datasets which deliberately pose complex DDS challenges for clustering. In order to improve the structure finding and clustering in high-dimensional DDS datasets, projection-based clustering (PBC) is introduced. The coexistence of projection and clustering allows to explore DDS through a topographic map. This enables to estimate, first, if any cluster tendency exists and, second, the estimation of the number of clusters. A comparison showed that PBC is always able to find the correct cluster structure, while the performance of the best of the 32 clustering algorithms varies depending on the dataset.

Algorithms implementing populations of agents which interact with one another and sense their environment may exhibit emergent behavior such as self-organization and swarm intelligence.Here a swarm system, called Databionic swarm (DBS), is introduced which is able to adapt itself to structures of high-dimensional data characterized by distance and/or density-based structures in the data space. By exploiting the interrelations of swarm intelligence, selforganization and emergence, DBS serves as an alternative approach to the optimization of a global objective function in the task of clustering. The swarm omits the usage of a global objective function and is parameter-free because it searches for the Nash equilibrium during its annealing process. To our knowledge, DBS is the first swarm combining these approaches. Its clustering can outperform common clustering methods such as K-means, PAM, single linkage, spectral clustering, model-based clustering, and Ward, if no prior knowledge about the data is available. A central problem in clustering is the correct estimation of the number of clusters. This is addressed by a DBS visualization called topographic map which allows assessing the number of clusters. It is known that all clustering algorithms construct clusters, irrespective of the data set contains clusters or not. In contrast to most other clustering algorithms, the topographic map identifies, that clustering of the data is meaningless if the data contains no (natural) clusters. The performance of DBS is demonstrated on a set of benchmark data, which are constructed to pose difficult clustering problems and in two real-world applications.

Biomedical data obtained during cell experiments, laboratory animal research, or human studies often display a complex distribution. Statistical identification of subgroups in research data poses an analytical challenge. Here were introduce an interactive R-based bioinformatics tool, called “AdaptGauss”. It enables a valid identification of a biologically-meaningful multimodal structure in the data by fitting a Gaussian mixture model (GMM) to the data. The interface allows a supervised selection of the number of subgroups. This enables the expectation maximization (EM) algorithm to adapt more complex GMM than usually observed with a noninteractive approach. Interactively fitting a GMM to heat pain threshold data acquired from human volunteers revealed a distribution pattern with four Gaussian modes located at temperatures of 32.3, 37.2, 41.4, and 45.4 °C. Noninteractive fitting was unable to identify a meaningful data structure. Obtained results are compatible with known activity temperatures of different TRP ion channels suggesting the mechanistic contribution of different heat sensors to the perception of thermal pain. Thus, sophisticated analysis of the modal structure of biomedical data provides a basis for the mechanistic interpretation of the observations. As it may reflect the involvement of different TRP thermosensory ion channels, the analysis provides a starting point for hypothesis-driven laboratory experiments.