Correlation detection method and connectivity-structure estimation method for neurons

An interneuron connection-structure estimation method for inferring a connection structure among three or more biological neurons interconnected by an unknown connection structure by calculating in a computer crossrelations among time-course data of activities of the three or more neurons, said interneuron connection-structure estimation method comprises the steps of:

• detecting firing activities of the three or more neurons;
• reading as time-course data the firing activities detected;
• computing in the computer conditional probabilities of one neuron's state or a combination of two neuron's states of the three or more neurons at a time difference under other neuron's states of the three or more neurons from said neuron-activity time-course data;
• computing in the computer conditional mutual information and multi-point mutual information used in the information theory from said computed conditional probabilities;
• estimating effects of third and subsequent neurons of the three or more neurons on a connectivity between first and second neurons of the three or more neurons;
• and
• inferring from the estimated effects a connection structure among the three or more neurons.

An interneuron connection-structure estimation method for inferring a connection structure among biological neurons by calculating a crossrelation among time-course data of activities of said neurons, said interneuron connectivity-structure estimation method comprises the steps of:

• detecting firing activities of the three or more neurons;
• reading as time-course data the firing activities detected;
• computing values of a conditional probability of a neuron's state from said neuron activity time-course data with time-course data of activities of another neuron taken as a condition;
• calculating a channel capacity defined as a maximum value of a mutual information with respect to firing probability of a presynaptic neuron from said computed conditional-probability values;
• distinguishing an inhibitory connectivity from an excitatory one by comparing said computed conditional-probability values to each other;
• and
• quantitatively estimating the magnitude of said crossrelation among said time-course data by using said channel capacity and distinction between excitatory and inhibitory connectivities.

An interneuron connection-structure estimation method for inferring a structure of connectivities among three or more biological neurons interconnected by an unknown connection structure by calculating a crossrelations among time-course data of activities of said three or more neurons, said interneuron connection-structure estimation method comprises the steps of:

• detecting firing activities of the three or more neurons;
• reading as time-course data the firing activities detected;
• computing values of a conditional probability of one neuron's state of the three or more neurons firing from said neuron activity time-course data with time-course data of activities of another neuron of the three or more neurons taken as a condition;
• calculating multi-point mutual information and conditional mutual information from said computed conditional-probability values;
• inferring an effect of a neuron of the three or more neurons on a connectivity between two other neurons of the three or more neurons by using said multi-point mutual information and said conditional mutual information;
• and
• inferring a connection structure among said three or more neurons from said effect and said multi-point mutual information quantities.

A method for preparing a neural network model comprising the steps of:

• determining a connection structure of a biological neural network on the basis of a connection structure estimation method;
• and
• preparing a neural network model on the basis of the determined connection structure;
• said connection structure estimation method comprising the steps of:
• calculating in a computer conditional probabilities of a neuron's state at a time difference under another neuron's state at a time difference under another neuron's state from said time-course data of said neuron activities;
• calculating in the computer a channel capacity from said calculated conditional probabilities, wherein said channel capacity is defined as a maximum value of mutual information, used in a well-known information theory, with respect to a firing probability of a presynaptic neuron;
• distinguishing an inhibitory connectivity from an excitatory connectivity by comparing in the computer said conditional probabilities to each other;
• quantitatively determining in the computer the magnitude of cross relation among said time-course data, on the basis of the channel capacity calculated;
• and
• using the magnitude of crossrelation determined to set the strength of the interneuron connectivity in the neural network model.

A method for preparing a neural network model comprising the steps of:

• determining a connection structure of a biological neural network on the basis of a connection structure estimation method;
• and
• preparing a neural network model on the basis of the determined connection structure;
• said connection structure estimation method comprising the steps of:
• detecting firing activities of three or more neurons;
• reading as time-course data the firing activities detected;
• computing in a computer conditional probabilities of one neuron's state or a combination of two neuron's state of the three or more neurons from said neuron-activity time-course data;
• computing in the computer conditional mutual information and multi-point mutual information, used in a well known information theory, from said computed conditional probabilities;
• estimating effects of third and subsequent neurons of the three or more neurons on a connectivity between first and second neurons of the three or more neurons;
• and
• inferring from the estimated effects a connection structure among the three or more neurons.

A neural network model implemented by a computer or on a neural network imitation circuit, prepared on the basis of a connection structure of a biological neural network determined by a connection structure estimation method,

• said connection structure estimation method comprising the steps of:
• calculating in a computer conditional probabilities of a neuron's state at a time difference under another neuron's state at a time difference under another neuron's state from said time-course data of said neuron activities;
• calculating in the computer a channel capacity from said calculated conditional probabilities, wherein said channel capacity is defined as a maximum value of mutual information, used in a well-known information theory, with respect to a firing probability of a presynaptic neuron;
• distinguishing an inhibitory connectivity from an excitatory connectivity by comparing in the computer said conditional probabilities to each other;
• quantitatively determining in the computer the magnitude of cross relation among said time-course data, on the basis of the channel capacity calculated;
• and
• using the magnitude of crossrelation determined to set the strength of the interneuron connectivity in the neural network model.

A neural network model implemented by a computer or on a neural network imitation circuit, prepared on the basis of a connection structure of a biological neural network determined by a connection structure estimation method,

• said connection structure estimation method comprising the steps of:
• detecting firing activities of three or more neurons;
• reading as time-course data the firing activities detected;
• computing in a computer conditional probabilities of one neuron's state or a combination of two neuron's state of the three or more neurons from said neuron-activity time-course data;
• computing in the computer conditional mutual information and multi-point mutual information, used in a well known information theory, from said computed conditional probabilities;
• estimating effects of third and subsequent neurons of the three or more neurons on a connectivity between first and second neurons of the three or more neurons;
• and
• inferring from the estimated effects a connection structure among the three or more neurons.