Results 1 -
2 of
2
Academic Publishers.
"... accessed by following the link in the citation at the bottom of the page. ..."
Abstract
- Add to MetaCart
accessed by following the link in the citation at the bottom of the page.
Temporal Segmentation with Bayesian Binning∗
"... Bayesian Binning (BB) is an exact inference technique which was originally de-veloped for applications in Computational Neuroscience, e.g. modeling spike count distributions [1] or estimating peri-stimulus time histograms (PSTH) [2]. BB encodes a (conditional) probability distribution (or density) w ..."
Abstract
- Add to MetaCart
(Show Context)
Bayesian Binning (BB) is an exact inference technique which was originally de-veloped for applications in Computational Neuroscience, e.g. modeling spike count distributions [1] or estimating peri-stimulus time histograms (PSTH) [2]. BB encodes a (conditional) probability distribution (or density) which is piece-wise constant in the domain of interest. This suggests that BB might be useful for retrospective temporal segmentation tasks, too. We illustrate the potential useful-ness of BB for temporal segmentation on two examples. First, we segment neural spike train data, demonstrating that BB is able to locate change points in the PSTH correctly. Second, we employ BB for (human) action sequence segmentation. We show that BB accurately identifies the transition points in the action sequence (e.g. a change from ’walking ’ to ’jumping’). 1