Resistive Memory Artificial Neuron for Excitatory Inhibitory Integration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional computing systems are inefficient in terms of power consumption and space due to the separation of external memory and processors, and existing artificial neurons fail to effectively emulate the excitatory and inhibitory updates of neuronal membrane potentials in neural networks.
Innovation Solution
An artificial neuron apparatus using two resistive memory cells, one for excitatory and one for inhibitory inputs, which change resistance in response to input signals and produce measurement signals that represent membrane potential differences, allowing for both excitatory and inhibitory updates and firing based on a threshold difference.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If hybrid analog/digital VLSI circuits are used to emulate neuronal functionality, then neuronal functionality can be emulated, but device complexity increases due to complex CMOS circuitry with large number of transistors
Solution Approach 1:
The patent replaces complex digital CMOS circuitry with resistive memory cells that inherently perform analog computation through their resistance values. The resistive memory cells directly emulate neuronal membrane potential integration and threshold firing behavior without requiring transistors or digital logic gates, thus substituting a complex mechanical/electronic system with a simpler physical phenomenon-based system.
Solution Approach 2:
The invention changes the operational parameter from digital voltage levels requiring multiple transistors to continuous resistance values of memristive devices. The resistive memory cells use their variable resistance state to represent membrane potential, eliminating the need for complex CMOS circuitry while maintaining neuronal computational functionality.
2Productivity
If conventional computing systems are used, then numerical calculations can be performed efficiently, but power consumption increases due to separation of external memory and processors
Solution Approach 1:
The patent merges the functions of memory storage and computational processing into a single resistive memory cell. The same device that stores information through its resistance state also performs the computational function of integrating input signals and determining when to fire, eliminating the need for separate memory and processor units and their associated data movement operations.
Solution Approach 2:
The resistive memory cell serves multiple functions simultaneously: it acts as a memory element storing the membrane potential state, an integrator summing excitatory and inhibitory inputs, a comparator detecting when threshold is reached, and a spike generator producing output signals. This multi-functionality eliminates the need for separate dedicated components for each function.
3Productivity
If conventional computing systems are used, then numerical calculations can be performed, but space requirements increase due to separation of external memory and processors
Solution Approach 1:
The patent combines memory storage and computational processing functions into a single resistive memory cell, eliminating the physical separation between memory and processor. This integration dramatically reduces the total area required by removing redundant interconnect structures, address decoders, and separate functional units that would be needed in a conventional von Neumann architecture.
4Ease of operation
If prior artificial neurons are used, then integrate-and-fire functionality can be emulated, but adaptability decreases as they cannot effectively handle both excitatory and inhibitory updates in neural networks
Solution Approach 1:
The patent segments the resistive memory cell into functionally distinct regions: an excitatory region that responds to excitatory inputs and an inhibitory region that responds to inhibitory inputs. Each region independently integrates its respective inputs and contributes to the overall membrane potential, enabling the neuron to simultaneously process both excitatory and inhibitory signals while maintaining the integrate-and-fire operation.
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 efficient and compact realization of artificial neurons capable of both excitatory and inhibitory updates, facilitating efficient operation in neural networks with reduced power consumption and improved integration in multi-neuron systems.
Implementation Method 1
Resistive memory cells are programmable-resistance devices which rely on the variable resistance characteristics of a volume of resistive material disposed between a pair of electrodes
Implementation Method 2
during said read phase, to apply a read current to their respective cells to produce first and second measurement signals respectively, dependent on resistance of the respective cells
Data Source
AI summary
Artificial neuron apparatus includes first and second resistive memory cells. The first resistive memory cell is connected in first circuitry having a first input and output. The second resistive memory cell is connected in second circuitry having a second input and output. The first and second circuitry are operable in alternating read and write phases to apply a programming current to their respective memory cells on receipt of excitatory and inhibitory neuron input signals, respectively. During the write phase, resistance of the respective cells is changed in response to successive excitatory and inhibitory neuron input signals. During the read phase, a read current is applied to their respective cells to produce first and second measurement signals, respectively. An output circuit connected to the first and second outputs produces a neuron output signal at a neuron output when a difference between the first and second measurement signals traverses a threshold.