The Experts below are selected from a list of 14457 Experts worldwide ranked by ideXlab platform
Alan L Yuille - One of the best experts on this subject based on the ideXlab platform.
-
unsupervised structure learning hierarchical recursive composition suspicious coincidence and Competitive Exclusion escholarship
2011Co-Authors: Long Zhu, Chenxi Lin, Haoda Huang, Yuanhao Chen, Alan L YuilleAbstract:We describe a new method for unsupervised structure learning of a hierarchical compositional model (HCM) for deformable objects. The learning is unsupervised in the sense that we are given a training dataset of images containing the object in cluttered backgrounds but we do not know the position or boundary of the object. The structure learning is performed by a bottom-up and top-down process. The bottom-up process is a novel form of hierarchical clustering which recursively composes proposals for simple structures to generate proposals for more complex structures. We combine standard clustering with the suspicious coincidence principle and the Competitive Exclusion principle to prune the number of proposals to a practical number and avoid an exponential explosion of possible structures. The hierarchical clustering stops automatically, when it fails to generate new proposals, and outputs a proposal for the object model. The top-down process validates the proposals and fills in missing elements. We tested our approach by using it to learn a hierarchical compositional model for parsing and segmenting horses on Weizmann dataset. We show that the resulting model is comparable with (or better than) alternative methods. The versatility of our approach is demonstrated by learning models for other objects (e.g., faces, pianos, butterflies, monitors, etc.). It is worth noting that the low-levels of the object hierarchies automatically learn generic image features while the higher levels learn object specific features.
-
unsupervised structure learning hierarchical recursive composition suspicious coincidence and Competitive Exclusion
European Conference on Computer Vision, 2008Co-Authors: Long Zhu, Chenxi Lin, Haoda Huang, Yuanhao Chen, Alan L YuilleAbstract:We describe a new method for unsupervised structure learning of a hierarchical compositional model (HCM) for deformable objects. The learning is unsupervised in the sense that we are given a training dataset of images containing the object in cluttered backgrounds but we do not know the position or boundary of the object. The structure learning is performed by a bottom-up and top-down process. The bottom-up process is a novel form of hierarchical clustering which recursively composes proposals for simple structures to generate proposals for more complex structures. We combine standard clustering with the suspicious coincidence principle and the Competitive Exclusion principle to prune the number of proposals to a practical number and avoid an exponential explosion of possible structures. The hierarchical clustering stops automatically, when it fails to generate new proposals, and outputs a proposal for the object model. The top-down process validates the proposals and fills in missing elements. We tested our approach by using it to learn a hierarchical compositional model for parsing and segmenting horses on Weizmann dataset. We show that the resulting model is comparable with (or better than) alternative methods. The versatility of our approach is demonstrated by learning models for other objects (e.g., faces, pianos, butterflies, monitors, etc.). It is worth noting that the low-levels of the object hierarchies automatically learn generic image features while the higher levels learn object specific features.
Long Zhu - One of the best experts on this subject based on the ideXlab platform.
-
unsupervised structure learning hierarchical recursive composition suspicious coincidence and Competitive Exclusion escholarship
2011Co-Authors: Long Zhu, Chenxi Lin, Haoda Huang, Yuanhao Chen, Alan L YuilleAbstract:We describe a new method for unsupervised structure learning of a hierarchical compositional model (HCM) for deformable objects. The learning is unsupervised in the sense that we are given a training dataset of images containing the object in cluttered backgrounds but we do not know the position or boundary of the object. The structure learning is performed by a bottom-up and top-down process. The bottom-up process is a novel form of hierarchical clustering which recursively composes proposals for simple structures to generate proposals for more complex structures. We combine standard clustering with the suspicious coincidence principle and the Competitive Exclusion principle to prune the number of proposals to a practical number and avoid an exponential explosion of possible structures. The hierarchical clustering stops automatically, when it fails to generate new proposals, and outputs a proposal for the object model. The top-down process validates the proposals and fills in missing elements. We tested our approach by using it to learn a hierarchical compositional model for parsing and segmenting horses on Weizmann dataset. We show that the resulting model is comparable with (or better than) alternative methods. The versatility of our approach is demonstrated by learning models for other objects (e.g., faces, pianos, butterflies, monitors, etc.). It is worth noting that the low-levels of the object hierarchies automatically learn generic image features while the higher levels learn object specific features.
-
unsupervised structure learning hierarchical recursive composition suspicious coincidence and Competitive Exclusion
European Conference on Computer Vision, 2008Co-Authors: Long Zhu, Chenxi Lin, Haoda Huang, Yuanhao Chen, Alan L YuilleAbstract:We describe a new method for unsupervised structure learning of a hierarchical compositional model (HCM) for deformable objects. The learning is unsupervised in the sense that we are given a training dataset of images containing the object in cluttered backgrounds but we do not know the position or boundary of the object. The structure learning is performed by a bottom-up and top-down process. The bottom-up process is a novel form of hierarchical clustering which recursively composes proposals for simple structures to generate proposals for more complex structures. We combine standard clustering with the suspicious coincidence principle and the Competitive Exclusion principle to prune the number of proposals to a practical number and avoid an exponential explosion of possible structures. The hierarchical clustering stops automatically, when it fails to generate new proposals, and outputs a proposal for the object model. The top-down process validates the proposals and fills in missing elements. We tested our approach by using it to learn a hierarchical compositional model for parsing and segmenting horses on Weizmann dataset. We show that the resulting model is comparable with (or better than) alternative methods. The versatility of our approach is demonstrated by learning models for other objects (e.g., faces, pianos, butterflies, monitors, etc.). It is worth noting that the low-levels of the object hierarchies automatically learn generic image features while the higher levels learn object specific features.
Christian C. Voigt - One of the best experts on this subject based on the ideXlab platform.
-
how bats escape the Competitive Exclusion principle seasonal shift from intraspecific to interspecific competition drives space use in a bat ensemble
Frontiers in Ecology and Evolution, 2018Co-Authors: Manuel Roeleke, Lilith Johannsen, Christian C. VoigtAbstract:Finding prey is crucial for predators that hunt on patchily distributed prey aggregations. At prey-rich patches, intraspecific and interspecific competition should be high. While the Competitive Exclusion principle suggests that species can only coexist if their ecological niches show considerable differences, newer theory suggests stabilizing and equalizing mechanisms besides classical niche differences that facilitate local coexistence. To identify such mechanisms, the understanding of the strength and nature (i.e. interference or exploitation) of competition in a species ensemble is a prerequisite. Here, we investigated intra- and interspecific competition between aerial-hawking insectivores, using the interactions between two open-space foraging bats as a model. In particular, we tested for shifts in space use of the common noctule bat Nyctalus noctula in response to simulated aggregations of conspecific and heterospecific competitors at foraging patches. When confronted with playbacks of heterospecific Pipistrellus nathusii, N. noctula increased their activity in the experimental area in early summer, but decreased activity in late summer. When confronted with playbacks of conspecifics, activity of N. noctula remained the same, irrespective of season. This pattern was accompanied by a decrease in the proportion of large insects during late summer. Our results suggest that intraspecific competition is more severe than interspecific competition for aerial insectivores in early summer. Probably, conspecifics engage in interference competition for flight space, and in the case of echolocating bats, may interfere with each other’s echolocation calls. Interspecific competition may be mediated by fine scale vertical partitioning and the use of different, non-interfering echolocation frequencies during insect rich times. In contrast, during late summer, bats may rather compete for the exploitation of relatively scarce large prey items. We speculate that N. noctula decreased activity in response to P. nathusii playbacks due to its inferior manoeuvrability and thus probably inferior hunting success in the presence of smaller, more agile bat species. However, N. noctula’s specialization on fast and efficient flight may enable them to use farther away and possibly less rich foraging patches, thus equalizing for a lower fitness compared to superior hunters.
-
Data_Sheet_2_How Bats Escape the Competitive Exclusion Principle—Seasonal Shift From Intraspecific to Interspecific Competition Drives Space Use in a Bat Ensemble.DOCX
2018Co-Authors: Manuel Roeleke, Lilith Johannsen, Christian C. VoigtAbstract:Predators that depend on patchily distributed prey face the problem of finding food patches where they can successfully compete for prey. While the Competitive Exclusion principle suggests that species can only coexist if their ecological niches show considerable differences, newer theory proposes that local coexistence can be facilitated by so-called stabilizing and equalizing mechanisms. A prerequisite to identify such mechanisms is the understanding of the strength and the nature of competition (i.e., interference or exploitation). We studied the interaction between two open-space foraging bats by testing if common noctule bats Nyctalus noctula shift their space use in response to simulated aggregations of conspecifics or heterospecific Pipistrellus nathusii. When confronted with playbacks of heterospecifics, N. noctula increased their activity in early summer, but decreased activity in late summer. This pattern was accompanied by a decrease in the proportion of large insects in late summer, suggesting a more intense competition for food in late compared to early summer. When confronted with playbacks of conspecifics, N. noctula did not change their activity, irrespective of season. Our results indicate that in early summer, intraspecific competition is more severe than interspecific competition for insectivorous bats. Likely, conspecifics engage in interference competition for flight space, and may suffer from reduced prey detectability as echolocation calls of conspecifics interfere with each other. During insect rich times, interspecific competition on the other hand may be mediated by fine scale vertical partitioning and the use non-interfering echolocation frequencies. In contrast, when food is scarce in late summer, bats may engage in exploitation competition. Our data suggests that N. noctula avoid aggregations of more agile bats like P. nathusii, probably due to impeded hunting success. Yet, as fast and efficient fliers, N. noctula may be able to escape this disadvantage by exploiting more distant foraging patches.
Yuanhao Chen - One of the best experts on this subject based on the ideXlab platform.
-
unsupervised structure learning hierarchical recursive composition suspicious coincidence and Competitive Exclusion escholarship
2011Co-Authors: Long Zhu, Chenxi Lin, Haoda Huang, Yuanhao Chen, Alan L YuilleAbstract:We describe a new method for unsupervised structure learning of a hierarchical compositional model (HCM) for deformable objects. The learning is unsupervised in the sense that we are given a training dataset of images containing the object in cluttered backgrounds but we do not know the position or boundary of the object. The structure learning is performed by a bottom-up and top-down process. The bottom-up process is a novel form of hierarchical clustering which recursively composes proposals for simple structures to generate proposals for more complex structures. We combine standard clustering with the suspicious coincidence principle and the Competitive Exclusion principle to prune the number of proposals to a practical number and avoid an exponential explosion of possible structures. The hierarchical clustering stops automatically, when it fails to generate new proposals, and outputs a proposal for the object model. The top-down process validates the proposals and fills in missing elements. We tested our approach by using it to learn a hierarchical compositional model for parsing and segmenting horses on Weizmann dataset. We show that the resulting model is comparable with (or better than) alternative methods. The versatility of our approach is demonstrated by learning models for other objects (e.g., faces, pianos, butterflies, monitors, etc.). It is worth noting that the low-levels of the object hierarchies automatically learn generic image features while the higher levels learn object specific features.
-
unsupervised structure learning hierarchical recursive composition suspicious coincidence and Competitive Exclusion
European Conference on Computer Vision, 2008Co-Authors: Long Zhu, Chenxi Lin, Haoda Huang, Yuanhao Chen, Alan L YuilleAbstract:We describe a new method for unsupervised structure learning of a hierarchical compositional model (HCM) for deformable objects. The learning is unsupervised in the sense that we are given a training dataset of images containing the object in cluttered backgrounds but we do not know the position or boundary of the object. The structure learning is performed by a bottom-up and top-down process. The bottom-up process is a novel form of hierarchical clustering which recursively composes proposals for simple structures to generate proposals for more complex structures. We combine standard clustering with the suspicious coincidence principle and the Competitive Exclusion principle to prune the number of proposals to a practical number and avoid an exponential explosion of possible structures. The hierarchical clustering stops automatically, when it fails to generate new proposals, and outputs a proposal for the object model. The top-down process validates the proposals and fills in missing elements. We tested our approach by using it to learn a hierarchical compositional model for parsing and segmenting horses on Weizmann dataset. We show that the resulting model is comparable with (or better than) alternative methods. The versatility of our approach is demonstrated by learning models for other objects (e.g., faces, pianos, butterflies, monitors, etc.). It is worth noting that the low-levels of the object hierarchies automatically learn generic image features while the higher levels learn object specific features.
Haoda Huang - One of the best experts on this subject based on the ideXlab platform.
-
unsupervised structure learning hierarchical recursive composition suspicious coincidence and Competitive Exclusion escholarship
2011Co-Authors: Long Zhu, Chenxi Lin, Haoda Huang, Yuanhao Chen, Alan L YuilleAbstract:We describe a new method for unsupervised structure learning of a hierarchical compositional model (HCM) for deformable objects. The learning is unsupervised in the sense that we are given a training dataset of images containing the object in cluttered backgrounds but we do not know the position or boundary of the object. The structure learning is performed by a bottom-up and top-down process. The bottom-up process is a novel form of hierarchical clustering which recursively composes proposals for simple structures to generate proposals for more complex structures. We combine standard clustering with the suspicious coincidence principle and the Competitive Exclusion principle to prune the number of proposals to a practical number and avoid an exponential explosion of possible structures. The hierarchical clustering stops automatically, when it fails to generate new proposals, and outputs a proposal for the object model. The top-down process validates the proposals and fills in missing elements. We tested our approach by using it to learn a hierarchical compositional model for parsing and segmenting horses on Weizmann dataset. We show that the resulting model is comparable with (or better than) alternative methods. The versatility of our approach is demonstrated by learning models for other objects (e.g., faces, pianos, butterflies, monitors, etc.). It is worth noting that the low-levels of the object hierarchies automatically learn generic image features while the higher levels learn object specific features.
-
unsupervised structure learning hierarchical recursive composition suspicious coincidence and Competitive Exclusion
European Conference on Computer Vision, 2008Co-Authors: Long Zhu, Chenxi Lin, Haoda Huang, Yuanhao Chen, Alan L YuilleAbstract:We describe a new method for unsupervised structure learning of a hierarchical compositional model (HCM) for deformable objects. The learning is unsupervised in the sense that we are given a training dataset of images containing the object in cluttered backgrounds but we do not know the position or boundary of the object. The structure learning is performed by a bottom-up and top-down process. The bottom-up process is a novel form of hierarchical clustering which recursively composes proposals for simple structures to generate proposals for more complex structures. We combine standard clustering with the suspicious coincidence principle and the Competitive Exclusion principle to prune the number of proposals to a practical number and avoid an exponential explosion of possible structures. The hierarchical clustering stops automatically, when it fails to generate new proposals, and outputs a proposal for the object model. The top-down process validates the proposals and fills in missing elements. We tested our approach by using it to learn a hierarchical compositional model for parsing and segmenting horses on Weizmann dataset. We show that the resulting model is comparable with (or better than) alternative methods. The versatility of our approach is demonstrated by learning models for other objects (e.g., faces, pianos, butterflies, monitors, etc.). It is worth noting that the low-levels of the object hierarchies automatically learn generic image features while the higher levels learn object specific features.