Cortical Spiking Signal Prediction via Point Process Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing models fail to accurately predict cortical spiking signals and functional links between cortical regions due to neglecting the point process features of neural signals, leading to challenges in model accuracy and prediction capability.
Innovation Solution
A method using a generalized linear model with Discrete Time Rescaling Kolmogorov-Smirnov Statistics as the optimization target and numerical gradient descent for optimizing parameters, incorporating natural features of point processes to improve prediction accuracy of cortical spiking neural signals, involving pretreatment, modeling, accuracy measurement, and iterative optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing models are used for prediction without incorporating point process features, then the model structure is simple, but the prediction accuracy of cortical spiking signals deteriorates
Solution Approach 1:
The patent changes the parameter of the optimization target from conventional mean squared error to Discrete Time Rescaling Kolmogorov-Smirnov Statistics, which specifically measures the distributional properties of point process data. This parameter change enables the model to capture the temporal statistics of neural spiking signals while maintaining the generalized linear model framework, thus improving prediction accuracy without excessive complexity increase
Solution Approach 2:
The patent substitutes the conventional optimization approach with numerical gradient descent method tailored for point process likelihood functions. This substitution replaces traditional optimization mechanics with a specialized algorithm that respects the discrete temporal nature of neural spiking data, improving model fitting accuracy while maintaining computational feasibility
2Reliability
If point process features are incorporated into the optimization target, then the prediction capability of neural spiking trains is improved, but the computational complexity increases
Solution Approach 1:
The patent performs preliminary discretization of the continuous neural spiking signals into discrete time bins before model fitting. This preliminary action transforms the continuous point process data into a format suitable for the generalized linear model with discrete optimization target, reducing the computational burden of handling continuous temporal data while preserving the essential spiking temporal structure for accurate prediction
Solution Approach 2:
The patent implements an iterative optimization process where the model parameters are refined through repeated cycles of prediction, error calculation using DTR-KS statistics, and parameter update via numerical gradient descent. This feedback mechanism progressively improves prediction capability by continuously adjusting parameters based on the discrepancy between predicted and actual spiking patterns, achieving high reliability through iterative refinement
Data Source
AI summary
The present invention is related to a method for prediction of cortical spiking neural signal, comprising the following steps: 1) pretreatment of spiking neural signal, 2) modeling of posterior cortical spiking neural signal generation probability, 3) model accuracy measurement, 4) model optimization with the help of numerical gradient descent and 5) prediction of posterior cortical spiking neural signal with iterative calculation. The prediction method aims to incorporate natural features of point process of spiking neural signal into the target for optimization of prediction model to improve model's capability in predicting neural spike trains.