Technical Program

Paper Detail

Paper: PS-1B.70
Session: Poster Session 1B
Location: H Fl├Ąche 1.OG
Session Time: Saturday, September 14, 16:30 - 19:30
Presentation Time:Saturday, September 14, 16:30 - 19:30
Presentation: Poster
Publication: 2019 Conference on Cognitive Computational Neuroscience, 13-16 September 2019, Berlin, Germany
Paper Title: Bayesian inference for an exploration-exploitation model of human gaze control
Manuscript:  Click here to view manuscript
License: Creative Commons License
This work is licensed under a Creative Commons Attribution 3.0 Unported License.
DOI: https://doi.org/10.32470/CCN.2019.1246-0
Authors: Noa Malem-Shinitski, Stefan Seelig, Sebastian Reich, Ralf Engbert, Potsdam University, Germany
Abstract: Understanding human gaze, and the saccadic selection process underlying it, is an important question in cognitive-neuroscience with many interesting applications in areas from psychology to computer vision. One way to advance our understanding is to develop generative models that capture the spatial interaction between fixations and the temporal structure of a sequence of fixations, known as scanpaths. Such models are scarce in the literature and even fewer attempt to model inter-subject variability. In this work, we present a new parametric model for scanpath generation. We develop a discrete-time probabilistic generative model, with a Markovian structure, where at each step the next fixation location is selected using one of two strategies - exploitation or exploration. We implement efficient Bayesian inference for hyperparameter estimation using an HMC within Gibbs approach. Our model is able to capture inter-observer variability in terms of saccade length and direction as demonstrated by fitting the model to a dataset of scanpaths from 35 subjects performing a task of free viewing of 30 natural scene image.