|
Model of Attention in Complex Natural Scene (Rolls - Deco)
|
Explains a model of attention interms of how objects and spatial attention (neural encoding, feature binding) operate in natural scenes. It claims Top down attention is much weaker in complex scenes, and that attention to details of an object in a natural scene is operated through the receptive fields of IT. The Inferior Temporal cortex receptive fields become small and located at the fovea to represent objects. Attention to features of objects is through competition and the respond to combinations of features is present in the correct relative spatial positions. Inferior temporal visual cortex neurons have asymmetrical receptive fields with respect to the fovea in complex scenes and this is how multiple Objects are perceived in a scene. The presented model uses an attractor network to demonstrate how the size of IT neuron changes based on scene complexity and results in different attractor states.
|
|
Predictive Coding for Action Understanding in the Mirror Neuron System (Friston - Mattout - Kilner)
|
A model of the mirror-neuron system (MNS) implementing a Bayes-optimal perception of actions emitted both by oneself or others. The premise of this article is that mirror neurons emerge naturally in any agent that acts on its environment to avoid surprising events. The use of predictive coding in conjunction with the MNS allows the model to predict the actions of self or others using proprioceptive signals. Ultimately, this supports the formulation of action as active inference.
|
|
Receptive fields of Inferior Temporal Visual Cortex in Natural Scenes (Trappenberg - Rolls - Stringer)
|
A model of the inferior temporal cortex focusing on receptive field size of IT neurons in natural scenes. The IT is modeled as an attractor network that learns translation invariant object representations. Cortical magnification, top-down bias, and scene complexity are shown to influence the effective size of IT receptive fields.
|
|
Visual attention using FEF feedback to enhance V4 (Hamker - Zirnsak)
|
A model of visual attention in simple scenes that uses feedback from frontal eye fields (FEF) to adjust the receptive field cells in V4. V4 is modeled as 3 separate layers for input, gain and pooling response from FEF feedback and projecting that information to IT. Pre-saccade feedback is incorporated into V4 to adjust receptive field.
|
