Memristive Nano-Strand Neural Hardware for Non-Deterministic Connectivity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional neuromorphic computing technologies face limitations in replicating the complex, non-deterministic connectivity and spike timing of biological neural networks, with existing silicon neuromorphic chips requiring digital interfaces and suffering from inefficiencies and deterministic connectivity patterns.
Innovation Solution
A neural network device comprising a mesh layer of randomly dispersed conductive nano-strands with memristor devices and modulating devices that mimic biological neural networks by enabling non-deterministic electrical communication between a large number of conductive nano-strands, allowing for dendritic-like connections and spiking neural network functionality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional silicon neuromorphic chips use digital interface and address-event representation protocol, then data exchange with remote memory is enabled, but computational efficiency deteriorates and deterministic connectivity patterns are imposed
Solution Approach 1:
The patent extracts the digital interface and address-event representation protocol from the neuromorphic computing system, replacing them with direct analog electrical signal transmission between conductive nano-strands. This removes the digital mediation layer that caused computational inefficiency while maintaining data exchange capability through physical electrical connections in the mesh network
Solution Approach 2:
The patent replaces the digital/software-based address-event representation protocol with a physical/electrical signal transmission system. Electrical signals propagate directly through the conductive nano-strand mesh without digital encoding/decoding, substituting a mechanical/physical transmission mechanism for the software-mediated digital protocol
2Manufacturing precision
If conventional neuromorphic chips use fixed connectivity patterns, then manufacturing precision is improved, but adaptability to biological neural network complexity deteriorates
Solution Approach 1:
The patent transforms the static, fixed connectivity pattern into a dynamic system where electrical signal propagation paths are determined by real-time signal characteristics, thresholds, and random walk processes. The effective connectivity emerges dynamically from the interaction of electrical signals with the conductive nano-strand mesh and modulating devices, rather than being predetermined by fixed physical connections
Solution Approach 2:
The patent changes the fundamental parameter of connectivity from fixed physical connections to variable electrical signal propagation characteristics. Factors such as signal amplitude, threshold values, and random displacement parameters determine effective connectivity, allowing the system to adapt its connectivity pattern to match biological neural network complexity
3Ease of operation
If conventional artificial neural networks use virtual mathematical entities, then software manipulation flexibility is improved, but physical implementation efficiency deteriorates
Solution Approach 1:
The patent replaces virtual mathematical neural network entities with physical electrical signal propagation through conductive nano-strands. Neural computation is implemented through actual electrical signal transmission, membrane potential changes, and physical threshold crossing events rather than software-based mathematical operations, achieving both physical implementation efficiency and operational flexibility
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The solution provides a physical neural network structure with enhanced capability and reliability for non-deterministic signal propagation, enabling a much larger number of interconnects beyond conventional methods, mimicking biological neural networks effectively.
Implementation Method 1
the modulating device is configured to automatically send an output signal to the electrode when stored potential at the modulating device due to the electrical signals exceeds a predetermined threshold value
Implementation Method 2
the memristor device is configured to send electrical signals received from the individual nano-strands of the first set of conductive nano-strands to the modulating device via the electrical conductor
Data Source
AI summary
Neural network systems and methods are provided. In one embodiment, a method of making a neural network device includes: forming a mesh layer on a substrate, the mesh layer including a matrix of randomly dispersed conductive nano-strands insulated from one another; forming an isolation trench extending into the mesh layer; forming a memristor device extending into the mesh layer, the memristor device including: an electrical conductor, and a layer of memristive material in electrical contact with individual nano-strands of a first set of conductive nano-strands in the mesh layer; forming an electrode extending into the mesh layer and spaced from the memristor device by the isolation trench, wherein the electrode is in electrical contact with individual conductive nano-strands of a second set of conductive nano-strands in the mesh layer; and forming a modulating device bridging the memristor device and the electrode.