Are redundant neurons redundant in categorical decision making? A robustness study of a network model

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Abstract

    In two-alternative forced-choice task paradigms, subjects can be required to make a categorical choice regardless of the difficulty of the task. This can be represented in attractor neural network models with an unstable steady state that separates the alternative choices. The unstable steady state can exist within a range of afferent inputs from outside the local circuit, but can be limited by physiological factors such as saturation of synapses or neuronal input-output function. The wider this decision-making "bandwidth" is, the more possible additional stimuli or cognitive controls within the brain can modulate a decision while the latter is still forming. In this work, we investigate this robustness problem in the context of neuronal diversity in the local network. More specifically, we want to understand how the decision-making bandwidth is affected by the number of redundant excitatory neurons which have activities uncorrelated with any choice outcome. By using a mean-field approach to study the stability of a spiking neuronal network model for decision making, we show that having a larger proportion of redundant neurons can reduce the decision-making bandwidth, and thus its robustness. Thus, we argue that there is a cost in having such neurons when operating a specific forced-choice task (exploitation), as opposed to their more obvious benefits of handling or learning multiple tasks (exploration). Finally, we also study the effects of synaptic or cellular heterogeneity on robust decision making.
    LanguageEnglish
    Title of host publicationUnknown Host Publication
    Number of pages1
    Publication statusPublished - 2009
    EventSociety for Neuroscience meeting 2009 - Washington D.C., USA
    Duration: 1 Jan 2009 → …

    Conference

    ConferenceSociety for Neuroscience meeting 2009
    Period1/01/09 → …

    Fingerprint

    Neurons
    Decision making
    Bandwidth
    Brain
    Neural networks
    Networks (circuits)
    Costs

    Cite this

    @inproceedings{7c6d06b4269d44589813a320abe1cc6c,
    title = "Are redundant neurons redundant in categorical decision making? A robustness study of a network model",
    abstract = "In two-alternative forced-choice task paradigms, subjects can be required to make a categorical choice regardless of the difficulty of the task. This can be represented in attractor neural network models with an unstable steady state that separates the alternative choices. The unstable steady state can exist within a range of afferent inputs from outside the local circuit, but can be limited by physiological factors such as saturation of synapses or neuronal input-output function. The wider this decision-making {"}bandwidth{"} is, the more possible additional stimuli or cognitive controls within the brain can modulate a decision while the latter is still forming. In this work, we investigate this robustness problem in the context of neuronal diversity in the local network. More specifically, we want to understand how the decision-making bandwidth is affected by the number of redundant excitatory neurons which have activities uncorrelated with any choice outcome. By using a mean-field approach to study the stability of a spiking neuronal network model for decision making, we show that having a larger proportion of redundant neurons can reduce the decision-making bandwidth, and thus its robustness. Thus, we argue that there is a cost in having such neurons when operating a specific forced-choice task (exploitation), as opposed to their more obvious benefits of handling or learning multiple tasks (exploration). Finally, we also study the effects of synaptic or cellular heterogeneity on robust decision making.",
    author = "KongFatt Wong-Lin",
    year = "2009",
    language = "English",
    booktitle = "Unknown Host Publication",

    }

    Wong-Lin, K 2009, Are redundant neurons redundant in categorical decision making? A robustness study of a network model. in Unknown Host Publication. Society for Neuroscience meeting 2009, 1/01/09.

    Are redundant neurons redundant in categorical decision making? A robustness study of a network model. / Wong-Lin, KongFatt.

    Unknown Host Publication. 2009.

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    TY - GEN

    T1 - Are redundant neurons redundant in categorical decision making? A robustness study of a network model

    AU - Wong-Lin, KongFatt

    PY - 2009

    Y1 - 2009

    N2 - In two-alternative forced-choice task paradigms, subjects can be required to make a categorical choice regardless of the difficulty of the task. This can be represented in attractor neural network models with an unstable steady state that separates the alternative choices. The unstable steady state can exist within a range of afferent inputs from outside the local circuit, but can be limited by physiological factors such as saturation of synapses or neuronal input-output function. The wider this decision-making "bandwidth" is, the more possible additional stimuli or cognitive controls within the brain can modulate a decision while the latter is still forming. In this work, we investigate this robustness problem in the context of neuronal diversity in the local network. More specifically, we want to understand how the decision-making bandwidth is affected by the number of redundant excitatory neurons which have activities uncorrelated with any choice outcome. By using a mean-field approach to study the stability of a spiking neuronal network model for decision making, we show that having a larger proportion of redundant neurons can reduce the decision-making bandwidth, and thus its robustness. Thus, we argue that there is a cost in having such neurons when operating a specific forced-choice task (exploitation), as opposed to their more obvious benefits of handling or learning multiple tasks (exploration). Finally, we also study the effects of synaptic or cellular heterogeneity on robust decision making.

    AB - In two-alternative forced-choice task paradigms, subjects can be required to make a categorical choice regardless of the difficulty of the task. This can be represented in attractor neural network models with an unstable steady state that separates the alternative choices. The unstable steady state can exist within a range of afferent inputs from outside the local circuit, but can be limited by physiological factors such as saturation of synapses or neuronal input-output function. The wider this decision-making "bandwidth" is, the more possible additional stimuli or cognitive controls within the brain can modulate a decision while the latter is still forming. In this work, we investigate this robustness problem in the context of neuronal diversity in the local network. More specifically, we want to understand how the decision-making bandwidth is affected by the number of redundant excitatory neurons which have activities uncorrelated with any choice outcome. By using a mean-field approach to study the stability of a spiking neuronal network model for decision making, we show that having a larger proportion of redundant neurons can reduce the decision-making bandwidth, and thus its robustness. Thus, we argue that there is a cost in having such neurons when operating a specific forced-choice task (exploitation), as opposed to their more obvious benefits of handling or learning multiple tasks (exploration). Finally, we also study the effects of synaptic or cellular heterogeneity on robust decision making.

    M3 - Conference contribution

    BT - Unknown Host Publication

    ER -