Spiking Neural Synapse Circuit for Excess Charge Transmission
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Spiking neural networks experience information missing and time delays due to excess charge not being transmitted to subsequent stages, leading to inaccurate image processing tasks.
Innovation Solution
A neural network device with synapse circuits that adjust synaptic current amplitude based on the excess charge component, using control signals to reflect this in the synaptic current output, thereby reducing information loss and transmission delay.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the charge amount held in the neuron circuit exceeds the threshold, then the spike signal is generated, but the excess charge amount is not transmitted to the subsequent stage causing information missing
Solution Approach 1:
A control signal output circuit is introduced as an intermediary element between the neuron circuit and synapse circuit. This control signal circuit extracts the excess charge component from the membrane potential and transmits it separately as a control signal to the synapse circuit, enabling the excess charge information to be conveyed without being lost in the threshold-based spike generation process.
Solution Approach 2:
The membrane potential signal is segmented into two separate components: the spike signal (voltage pulse) that indicates neuronal firing, and the control signal that carries the excess charge information. This segmentation allows both the threshold-based activation and the excess charge magnitude to be transmitted independently to the synapse circuit, preventing information loss.
2Productivity
If the reset discharge amount is adjusted using the excess charge amount, then the occurrence frequency of spike signal is increased, but a time delay occurs in reflecting the excess charge amount
Solution Approach 1:
The control signal output circuit generates the control signal representing the excess charge component immediately when the membrane potential exceeds the threshold, before the neuron circuit undergoes reset discharge. This preliminary generation of the control signal ensures that the excess charge information is captured and transmitted without delay, independent of the subsequent reset operation.
Solution Approach 2:
The control signal acts as an intermediary that carries the excess charge information from the neuron circuit to the synapse circuit in real-time, bypassing the time delay inherent in adjusting reset discharge amounts after spike generation. This separate signal pathway enables immediate reflection of excess charge magnitude in the synaptic current.
3Use of energy by moving object
If discrete information processing is used in spiking neural network, then energy consumption is reduced, but information missing occurs during image processing tasks
Solution Approach 1:
The invention changes the parameter representation by encoding not only the presence/absence of spike events but also the magnitude of excess charge as a separate control signal parameter. This dual-parameter approach (spike timing + excess charge magnitude) enriches the discrete spiking representation with additional information while maintaining the energy-efficient spiking paradigm.
Solution Approach 2:
The control signal serves as an intermediary that bridges the gap between discrete spike events and continuous information representation needed for image processing. By transmitting the excess charge magnitude as a separate analog or digital control signal, the system preserves continuous information content while utilizing energy-efficient discrete spike-based communication for neuronal activation.
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 device achieves accurate spiking-type neural networks with reduced information loss and faster processing times, enabling efficient image recognition and classification without requiring CPUs or GPUs.
Implementation Method 1
a charge holding circuit configured to accumulate charge corresponding to the synaptic current acquired by the input circuit and generate a membrane potential corresponding to the accumulated charge
Implementation Method 2
outputs a synaptic current obtained by multiplying the spike signal by a synaptic weight by a variable resistance element or the like
Data Source
AI summary
A neural network device according to an embodiment includes a plurality of synapse circuits and a plurality of neuron circuits. A first neuron circuit out of the neuron circuits includes an input circuit, a charge holding circuit, a comparison circuit, a firing circuit, a charge control circuit, and a control signal output circuit. When a determination signal changes from a second value to a first value, the firing circuit outputs a spike signal. The control signal output circuit outputs a control signal indicating a comparison voltage that is based on an excess component of a membrane potential exceeding a threshold potential. In response to acquiring the spike signal from the first neuron circuit, a first synapse circuit out of the synapse circuits that acquires the spike signal from the first neuron circuit outputs a synaptic current of a current amount corresponding to the control signal and a synaptic weight.