Anticipatory models in gaze control: a developmental model

Infants' saccadic behavior during 2-dimensional displays of a bounce

The study examined the saccadic behavior of 4- to 10-month-old infants when tracking a two-dimensional linear motion of a circle that occasionally bounced off a barrier constituted by the screen edges. It was investigated whether infants could anticipate the angle of the circle's direction after the bounce and the circle's displacement from the location of bounce. Seven bounce types were presented which differed in the angle of incidence. Three of the seven bounce types showed physically implausible bounces. Saccades that started before the infant could perceive the bounce and ended after the bounce were analyzed. Infants' saccades matched the circle's displacement with sufficient accuracy to conclude that they made predictive saccades. Only results from two bounce types where the circle's speed was less than 12.5°/s allowed for the possibility that infants made reactive saccades. The infants' anticipated angle was close to the angle of the circle's direction after the bounce. When the circle was moving at 40°/s, the difference between the two aforementioned angles was less than 15°, but it increased as the circle became slower. The effect of age on the saccade targets and other object-tracking measures was small and mostly masked by a large estimation error. Estimates of the saccade amplitude, saccade frequency, and the gaze-circle displacement were similar to those observed for saccades made when no bounce occurred and they were also similar to those reported in previous studies of infant object tracking with similar trajectories but without a barrier.

Ikaros: A framework for controlling robots with system-level brain models

Ikaros is an open framework for system-level brain modeling and real-time robot control. Version 2 of the system includes a range of computational components that implements various algorithms and methods ranging from models of neural circuits to control systems and hardware interfaces for robot. Ikaros supports the design and implementation of large-scale computation models using a flow programming paradigm. Version 2 includes a number of new features that support complex networks of hierarchically arranged components as well as a web-based interactive editor. More than 100 persons have contributed to the code base and over 100 scientific publications report on work that has used Ikaros for simulations or robot control.

Toward a Unified Sub-symbolic Computational Theory of Cognition

This paper proposes how various disciplinary theories of cognition may be combined into a unifying, sub-symbolic, computational theory of cognition. The following theories are considered for integration: psychological theories, including the theory of event coding, event segmentation theory, the theory of anticipatory behavioral control, and concept development; artificial intelligence and machine learning theories, including reinforcement learning and generative artificial neural networks; and theories from theoretical and computational neuroscience, including predictive coding and free energy-based inference. In the light of such a potential unification, it is discussed how abstract cognitive, conceptualized knowledge and understanding may be learned from actively gathered sensorimotor experiences. The unification rests on the free energy-based inference principle, which essentially implies that the brain builds a predictive, generative model of its environment. Neural activity-oriented inference causes the continuous adaptation of the currently active predictive encodings. Neural structure-oriented inference causes the longer term adaptation of the developing generative model as a whole. Finally, active inference strives for maintaining internal homeostasis, causing goal-directed motor behavior. To learn abstract, hierarchical encodings, however, it is proposed that free energy-based inference needs to be enhanced with structural priors, which bias cognitive development toward the formation of particular, behaviorally suitable encoding structures. As a result, it is hypothesized how abstract concepts can develop from, and thus how they are structured by and grounded in, sensorimotor experiences. Moreover, it is sketched-out how symbol-like thought can be generated by a temporarily active set of predictive encodings, which constitute a distributed neural attractor in the form of an interactive free-energy minimum. The activated, interactive network attractor essentially characterizes the semantics of a concept or a concept composition, such as an actual or imagined situation in our environment. Temporal successions of attractors then encode unfolding semantics, which may be generated by a behavioral or mental interaction with an actual or imagined situation in our environment. Implications, further predictions, possible verification, and falsifications, as well as potential enhancements into a fully spelled-out unified theory of cognition are discussed at the end of the paper.

Simulating the role of visual selective attention during the development of perceptual completion

We recently proposed a multi-channel, image-filtering model for simulating the development of visual selective attention in young infants (Schlesinger, Amso & Johnson, 2007). The model not only captures the performance of 3-month-olds on a visual search task, but also implicates two cortical regions that may play a role in the development of visual selective attention. In the current simulation study, we used the same model to simulate 3-month-olds' performance on a second measure, the perceptual unity task. Two parameters in the model - corresponding to areas in the occipital and parietal cortices - were systematically varied while the gaze patterns produced by the model were recorded and subsequently analyzed. Three key findings emerged from the simulation study. First, the model successfully replicated the performance of 3-month-olds on the unity perception task. Second, the model also helps to explain the improved performance of 2-month-olds when the size of the occluder in the unity perception task is reduced. Third, in contrast to our previous simulation results, variation in only one of the two cortical regions simulated (i.e. recurrent activity in posterior parietal cortex) resulted in a performance pattern that matched 3-month-olds. These findings provide additional support for our hypothesis that the development of perceptual completion in early infancy is promoted by progressive improvements in visual selective attention and oculomotor skill.

Anticipation requires adaptation

To successfully interact with a dynamic world, our actions must be guided by a continuously changing anticipated future. Such anticipations must be tuned to the processing delays in the nervous system as well as to the slowness of the body, something that requires constant adaptation of the predictive mechanisms, which in turn require that sensory information be processed at different time-scales.