Christoph Schnörr | Self-Assignment Flows For Unsupervised Contextual Data Labeling
Автор: CEA
Загружено: 2020-03-30
Просмотров: 252
Описание:
Watch Christoph Schnörr's talk during the First French-German Meeting in Physics, Mathematics and Artificial Intelligence Theory that took place from November 4 to 6, 2019 in Paris.
Abstract:
Self-assignment flows provide a specific answer to the problem to design a mathematical and algorithmic approach that seamlessly combines the computation of data prototypes for data coding and data partitioning using these prototypes. The data are assumed to be given on an arbitrary graph with affinities as edge weights. Scale in terms of the graph distance between vertices is the only parameter that determines what structure in the data is detected in this way. By construction, assignment flows at coarser scales have the same mathematical form and, therefore, may contribute to the understanding of deep networks in the long run. The talk concludes with a brief outlook along this line of research, from the viewpoint of the STRUCTURES project at Heidelberg.
_____
Follow us on:
LinkedIn: / cealist
Twitter: / cea_list
Website: http://www-list.cea.fr/en
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке:
