Yu, Yingwei (2006-08). Computational role of disinhibition in brain function. Doctoral Dissertation. Thesis uri icon

abstract

  • Neurons are connected to form functional networks in the brain. When neurons are combined in sequence, nontrivial effects arise. One example is disinhibition; that is, inhibition to another inhibitory factor. Disinhibition may be serving an important purpose because a large number of local circuits in the brain contain disinhibitory connections. However, their exact functional role is not well understood. The objective of this dissertation is to analyze the computational role of disinhibition in brain function, especially in visual perception and attentional control. My approach is to propose computational models of disinhibition and then map the model to the local circuits in the brain to explain psychological phenomena. Several computational models are proposed in this dissertation to account for disinhibition. (1) A static inverse difference of Gaussian filter (IDoG) is derived to account explicitly for the spatial effects of disinhibition. IDoG can explain a number of complex brightness-contrast illusions, such as the periphery problem in the Hermann grid and the White's effect. The IDoG model can also be used to explain orientation perception of multiple lines as in the modified version of Poggendorff illusion. (2) A spatio-temporal model (IDoGS) in early vision is derived and it successfully explains the scintillating grid illusion, which is a stationary display giving rise to a striking, dynamic, scintillating effect. (3) An interconnected Cohen-Grossberg neural network model (iCGNN) is proposed to address the dynamics of disinhibitory neural networks with a layered structure. I derive a set of sufficient conditions for such an interconnected system to reach asymptotic stability. (4) A computational model combining recurrent and feed-forward disinhibition is designed to account for input-modulation in temporal selective attention. The main contribution of this research is that it developed a unified framework of disinhibition to model several different kinds of neural circuits to account for various perceptual and attentional phenomena. Investigating the role of disinhibition in the brain can provide us with a deeper understanding of how the brain can give rise to intelligent and complex functions.
  • Neurons are connected to form functional networks in the brain. When neurons are
    combined in sequence, nontrivial effects arise. One example is disinhibition; that is,
    inhibition to another inhibitory factor. Disinhibition may be serving an important
    purpose because a large number of local circuits in the brain contain disinhibitory
    connections. However, their exact functional role is not well understood.
    The objective of this dissertation is to analyze the computational role of disinhibition
    in brain function, especially in visual perception and attentional control.
    My approach is to propose computational models of disinhibition and then map the
    model to the local circuits in the brain to explain psychological phenomena. Several
    computational models are proposed in this dissertation to account for disinhibition.
    (1) A static inverse difference of Gaussian filter (IDoG) is derived to account explicitly
    for the spatial effects of disinhibition. IDoG can explain a number of complex
    brightness-contrast illusions, such as the periphery problem in the Hermann grid and
    the White's effect. The IDoG model can also be used to explain orientation perception
    of multiple lines as in the modified version of Poggendorff illusion. (2) A
    spatio-temporal model (IDoGS) in early vision is derived and it successfully explains
    the scintillating grid illusion, which is a stationary display giving rise to a striking,
    dynamic, scintillating effect. (3) An interconnected Cohen-Grossberg neural network
    model (iCGNN) is proposed to address the dynamics of disinhibitory neural networks with a layered structure. I derive a set of sufficient conditions for such an interconnected
    system to reach asymptotic stability. (4) A computational model combining
    recurrent and feed-forward disinhibition is designed to account for input-modulation
    in temporal selective attention.
    The main contribution of this research is that it developed a unified framework of
    disinhibition to model several different kinds of neural circuits to account for various
    perceptual and attentional phenomena. Investigating the role of disinhibition in the
    brain can provide us with a deeper understanding of how the brain can give rise to
    intelligent and complex functions.

publication date

  • August 2006