Neural Network Apparatus With Secondary Battery Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional neuromorphic processors simulating spike-timing-dependent plasticity (STDP) require additional feedback paths, increasing circuit size and power consumption, and lack energy sources similar to human neurons.
Innovation Solution
A neural network apparatus with neuron circuits that include integration, firing, and secondary battery components, allowing energy storage and feedback of firing timing without external feedback paths, using resistance change memory elements and switches to update synaptic coefficients based on pulse signals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If feedback paths are added to simulate STDP in conventional neuromorphic processors, then learning capability is improved, but circuit size increases
Solution Approach 1:
The patent combines the feedback path functionality directly into the neuron circuit by integrating a secondary battery that automatically provides feedback based on firing timing. This merging eliminates the need for separate feedback paths while maintaining STDP learning capability, thus improving adaptability without increasing circuit size.
Solution Approach 2:
The secondary battery serves multiple functions: it provides power to the neuron circuit and simultaneously provides feedback signals for STDP learning based on firing timing. This multi-functionality allows the same component to achieve both power supply and learning feedback, reducing overall circuit complexity while maintaining learning capability.
2Adaptability or versatility
If feedback paths are added to simulate STDP in conventional neuromorphic processors, then learning capability is improved, but power consumption increases
Solution Approach 1:
The patent merges the power supply and feedback functions into a single secondary battery component. This integration allows the feedback mechanism to operate passively using the battery's natural voltage changes during firing, eliminating the need for additional active feedback circuits that would consume extra power, thus reducing overall power consumption while maintaining learning capability.
Solution Approach 2:
The secondary battery automatically provides feedback signals based on its own voltage changes during neuron firing without requiring external control circuits. This self-service mechanism eliminates the need for additional power-consuming feedback path components, achieving STDP learning with minimal power consumption.
3Adaptability or versatility
If conventional neuromorphic processors use external feedback paths, then STDP learning is achieved, but energy efficiency decreases
Solution Approach 1:
The secondary battery automatically generates feedback signals based on its own voltage changes during firing events, without requiring external feedback paths or additional energy sources. This self-service mechanism achieves STDP learning while minimizing energy loss by using the battery's natural operational characteristics for feedback generation.
Solution Approach 2:
The patent implements feedback through the secondary battery's voltage changes that naturally occur during firing. This feedback mechanism is integrated into the power supply system itself, creating an energy-efficient closed-loop system where the feedback path is formed by the battery's inherent electrical characteristics rather than separate feedback circuits.
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 apparatus efficiently simulates STDP learning by updating synaptic strengths based on neuron firing timing with reduced circuit complexity and power consumption, mimicking human neuronal energy generation and feedback mechanisms.
Implementation Method 1
The secondary battery is configured to supply the firing circuit with drive electric power used for generating the pulse signal
Implementation Method 2
The integration circuit is configured to output an integral signal obtained by integrating input signals
Implementation Method 3
The firing circuit is configured to generate, in accordance with the integral signal, a pulse signal to be transmitted to the neuron circuit provided at a subsequent layer
Implementation Method 4
using resistance change memory elements and switches to update synaptic coefficients based on pulse signals
Data Source
AI summary
According to an embodiment, a neural network apparatus includes a plurality of neuron circuits, each including an integration circuit, a firing circuit, and a secondary battery. The integration circuit is configured to output an integral signal obtained by integrating input signals. The firing circuit is configured to generate, in accordance with the integral signal, a pulse signal to be transmitted to the neuron circuit provided at a subsequent layer. The secondary battery is configured to supply the firing circuit with drive electric power used for generating the pulse signal.