Neuromorphic System Bias Current Switching

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional neuromorphic systems face challenges in efficiently switching between multiple functional operations due to the von Neumann bottleneck, which limits their ability to process large amounts of data in real-time and requires significant memory and power consumption when performing multi-functional tasks.

Innovation Solution

A neuromorphic system with a controlling unit and a neuron unit that regulates bias currents to control the excitatory and inhibitory neuron groups, allowing the system to switch between high-activity, middle-activity, and low-activity states, enabling flexible execution of various functions within the same circuit without changing connection strengths.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If an artificial neural network is designed to perform multiple functional tasks, then the system versatility is improved, but the memory consumption and power consumption increase significantly

Engineering Contradiction:
Improvefunctional operation versatilityVSAvoidpower consumption
Core Design Contradiction:
Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent applies parameter changes by dynamically adjusting the bias current injected into each neuron to switch between different functional operations. Instead of reconfiguring connection strengths or adding hardware components, the system changes the operational parameters (bias current levels) to transition between high-activity, middle-activity, and low-activity states, enabling the same neural network to perform multiple functions with constant memory and power consumption.

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If an artificial neural network is designed to perform multiple functional tasks, then the system versatility is improved, but the device complexity increases

Engineering Contradiction:
Improvefunctional operation versatilityVSAvoidsystem complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent implements universality by designing a single neural network that can perform multiple functional operations through dynamic bias current control. The same physical hardware (neurons, synapses, and connections) is used for different functions by simply adjusting the bias current parameters, eliminating the need for separate circuits or reconfigurable components for each function.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Reliability

If conventional computer architecture is used for large amount of data processing, then the system stability is maintained, but the data processing speed is limited by the von Neumann bottleneck

Engineering Contradiction:
Improvesystem stabilityVSAvoiddata processing speed
Core Design Contradiction:
ReliabilityVSSpeed

Solution Approach 1:

The patent replaces the conventional von Neumann architecture with a neuromorphic system that uses continuous analog signals and parallel processing. Instead of discrete digital operations with sequential data movement between memory and processor, the system uses continuous membrane potential changes and simultaneous neural computations, achieving real-time processing while maintaining stability through the inherent robustness of neural network operations.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS12079709B2Neuromorphic system and methodology for switching between functional operations
Publication Date: 2024.09.03 NATIONAL TSING HUA UNIVERSITY
  • US12079709B2 patent drawing
  • US12079709B2 patent drawing
  • US12079709B2 patent drawing

AI summary

A neuromorphic system for switching between a multitude of functional operations includes a controlling unit and a neuron unit. The controlling unit provides a first input and a second input and regulates a multitude of bias currents. The neuron unit receives the bias currents. An input neuron group receives the first input and the second input. An excitatory neuron group is stimulated by the input neuron group. An inhibitory neuron group is electrically connected to the excitatory neuron group, and the inhibitory neuron group and the excitatory neuron group stimulate each other. An output neuron is electrically connected to the excitatory neuron group and stimulated by the excitatory neuron group to generate an output. The bias currents control the excitatory neuron group, the inhibitory neuron group and the output neuron to be in one of a high-activity state, a middle-activity state and a low-activity state.