Correlated Electron Switch Elements for Neuromorphic Computing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current artificial neural networks lack efficient mechanisms for storing synaptic weights and transitioning between impedance states quickly, which limits their ability to mimic biological neurons and synapses effectively.
Innovation Solution
The use of correlated electron switch (CES) elements, which can switch between high and low impedance states based on reset and set conditions, allowing for the accumulation and storage of currents and voltages to produce signals when thresholds are met, and can be configured to function as either neurons or synapses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If conventional artificial neural networks are used, then basic neural network functions can be performed, but the ability to store synaptic weights and transition between impedance states quickly is limited
Solution Approach 1:
The patent applies parameter changes by utilizing the resistive switching characteristics of correlated electron materials to transition between distinct impedance states (high and low resistance states). These state transitions enable fast synaptic weight storage and retrieval, directly addressing the contradiction between transition speed and storage reliability. The material's ability to maintain stable resistance states ensures reliable weight storage while the abrupt transitions provide high speed.
Solution Approach 2:
The patent employs composite material structures combining correlated electron materials with electrode layers to create functional neural network components. This composite approach leverages the unique electronic correlations in the material to achieve both fast switching speeds and reliable weight storage, resolving the technical contradiction between speed and reliability in synaptic weight management.
2Productivity
If conventional neural networks are used, then basic processing can be performed, but efficient accumulation and storage of inputs is limited
Solution Approach 1:
The patent merges multiple functions into the correlated electron switch element: it simultaneously performs input accumulation through conductance modulation, synaptic weight storage through resistive states, and signal transmission through impedance transitions. This consolidation eliminates sequential processing delays and improves overall productivity while reducing time loss in neural network operations.
Solution Approach 2:
The patent enables continuous accumulation of input signals through the persistent conductance changes in correlated electron materials. The material maintains accumulated states without requiring continuous power supply, allowing uninterrupted signal integration and processing, thereby improving productivity while minimizing time loss.
3Measurement precision
If conventional neural networks are used, then basic signal transmission can be achieved, but precise signal transmission and processing is limited
Solution Approach 1:
The patent implements universal neural network components where the correlated electron switch element can function as both synapse (for weight storage) and neuron (for signal accumulation and transmission). This multi-functionality reduces device complexity by eliminating the need for separate components while maintaining precise signal transmission through the material's well-defined impedance states.
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
Enables fast and efficient storage of synaptic weights and accumulation of inputs, facilitating precise signal transmission and processing, with the ability to quickly transition between states, thereby enhancing the performance of artificial neural networks.
Implementation Method 1
an abrupt conductive or insulative state transition arising from electron correlations rather than solid state structural phase changes
Data Source
AI summary
Broadly speaking, the present techniques exploit the properties of correlated electron materials for artificial neural networks and neuromorphic computing. In particular, the present techniques provide apparatuses/devices that comprise at least one correlated electron switch (CES) element and which may be used as, or to form, an artificial neuron or an artificial synapse.