Neural Network Map Construction from Membrane Potentials
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current neuromorphic engineering methods face challenges in accurately mapping the connections and synaptic weights of natural neural networks due to limitations in measuring membrane potentials, particularly with extracellular electrodes, which result in noisy and distorted signals, making it difficult to replicate the structure and function of biological neural networks in electronic devices.
Innovation Solution
A method is developed to construct a neural network map based on membrane potentials of biological neurons using intracellular electrodes, identifying connection structures and estimating synaptic weights, and mapping these to electronic neural networks, allowing for the reproduction of natural neural network operations in electronic devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If extracellular electrodes are used to measure membrane potentials, then the measurement process is simpler and less invasive, but the signal quality deteriorates with noise and distortion
Solution Approach 1:
The patent uses an intermediary processing system that includes a measurement unit, a determination unit, and a mapping unit. The determination unit acts as an intermediary that processes the raw extracellular signals to identify action potentials and post-synaptic potentials, separating useful information from noise. This intermediary processing layer enables the system to maintain ease of measurement with extracellular electrodes while achieving accurate neural network mapping.
2Device complexity
If extracellular electrodes are used to measure membrane potentials, then the device complexity is reduced, but the mapping accuracy of neural connections deteriorates
Solution Approach 1:
The patent segments the measurement and processing functions into distinct modular units: a measurement unit for acquiring signals, a determination unit for identifying specific potential types (action potentials and post-synaptic potentials), and a mapping unit for constructing the neural network map. This segmentation allows the system to maintain low device complexity while achieving high mapping accuracy through specialized processing of different signal components.
Solution Approach 2:
The patent creates a digital copy or model of the neural network connections by mapping the measured extracellular signals to represent actual neural pathways and synaptic weights. Instead of requiring direct intracellular measurement of each connection, the system copies the functional relationships from the extracellular measurements to construct an accurate neural network map, achieving high mapping accuracy without proportionally increasing device complexity.
3Productivity
If limited numbers of biological neurons are targeted for measurement, then the measurement process becomes more manageable, but the comprehensiveness of neural network mapping deteriorates
Solution Approach 1:
The patent employs extracellular electrodes that can simultaneously measure signals from multiple neurons and multiple types of potentials (action potentials and post-synaptic potentials) using the same measurement system. This multi-functional capability allows the system to maintain high productivity with a single electrode setup while achieving comprehensive neural network mapping by capturing diverse neural activities from multiple neurons concurrently.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Disclosed is an apparatus and method mapping a natural neural network into an electronic neural network device of an electronic device. The method includes constructing a neural network map of a natural neural network based on membrane potentials of a plurality of biological neurons of the natural neural network, where the membrane potentials correspond to at least two different respective forms of membrane potentials, and mapping the neural network map to the electronic neural network device. The constructing of the neural network map and the mapping of the neural network map implement learning of the electronic neural network device. The method may further includes obtaining an input or stimuli, activating the learned electronic neural network device, provided the obtained input or stimuli, to perform neural network operations, and generating a neural network result for the obtained input or stimuli based on a result of the activated learned electronic neural device.