English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

A fundamental aspect of human intelligence is our ability to express and understand relations. For example, we can easily find an object described by the phrase “the red car to the left of the silver car”. Just as easily, we can describe a scene ourselves and say something like “the red car is moving toward the intersection”. e processes in our brain that enable us to interpret such spatial relations and movement relations in the world may also facilitate the processing of a much richer body of abstract relations that show up most prominently in language. Understanding these processes is thus fundamental to understanding the link between cognition and language. To capture how spatial and movement relations may be resolved, this thesis introduces a neural process model based on dynamic field theory (DFT), a mathematical and conceptual framework for modeling cognitive processes. e model receives camera input showing a white table with colored balls on it as well as input that represents a relational phrase, similar to the examples above. e task of the model is to bring the described object into the attentional foreground—to ground the phrase. In another type of task, it must generate a phrase itself, describing the visual scene. Solving these tasks requires that the model is able to map discrete concepts in the phrase to continuous feature dimensions, sequentially guide attentional processes to search the scene for multiple objects, put them in working memory, adjust the reference frame of their representation, and evaluate their fit with representations of spatial relations and movement relations. A particular challenge and key aspect of this thesis is to have the model organize its own processes and perform the above tasks without human interference. Establishing the model based on neural principles additionally requires solving fundamental neural problems, such as the neural pointer problem, the binding problem, and generating discrete processing steps from processes that evolve in continuous time. e model solves all of these problems and innovates over previous work by capturing both grounding and description tasks, spatial and movement relations, and a flexible, hierarchical organization of its processes. is is demonstrated in 104 qualitatively different tests that vary the task, the configuration of objects, and, in particular, how well the given phrase matches the scene. e model is able to correctly ground the given phrase, or generate a phrase, in all cases where this is possible. By demonstrating how spatial and movement relations may be captured by a neural process model, the thesis brings DFT one step closer toward a comprehensive neural theory of cognition.

A neural dynamic model for the perceptual grounding of spatial and movement relations