Post-Synaptic Inverting Circuits for Smooth Neuromorphic Transfer Curves
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current neuromorphic devices lack the ability to produce smooth output signal curves, which limits their capability to represent multiple levels of synaptic weights and input data patterns effectively.
Innovation Solution
The neuromorphic device incorporates a post-synaptic neuron with a transfer function circuit that includes a series connection of inverting circuits with varying numbers and types of pull-up and pull-down transistors, allowing for smooth output signal changes resembling a sigmoid curve, thereby enabling multiple levels of signal representation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional neuromorphic device circuits are used, then device complexity is reduced, but the ability to produce smooth output signal curves is lost
Solution Approach 1:
The transfer function circuit is segmented into multiple inverting circuits connected in series, where each inverting circuit contains pull-up and pull-down transistor groups. This segmentation allows the complex transfer function to be achieved through modular circuit blocks, resolving the contradiction by breaking down the complex function into manageable segments that collectively produce smooth output curves.
Solution Approach 2:
Different regions of the transfer function circuit exhibit different characteristics through varying numbers of pull-up and pull-down transistors in different inverting circuits. This local quality variation enables the circuit to produce smooth sigmoidal output curves in critical regions while maintaining manageable complexity elsewhere, as each local section contributes differently to the overall transfer function.
2Loss of information
If simple transfer function circuits are used, then device complexity is reduced, but the representation of multiple synaptic weight levels is limited
Solution Approach 1:
The transfer function circuit employs dynamically controllable transistor configurations where the effective number of pull-up and pull-down transistors can be adjusted through gating mechanisms. This dynamic control enables the circuit to represent multiple synaptic weight levels by modulating the transfer function characteristics, achieving high information representation without requiring a permanently complex circuit structure.
Solution Approach 2:
The circuit achieves multiple synaptic weight levels by changing operational parameters such as the effective transistor count and conductivity ratios rather than requiring physically different circuit topologies. By varying parameters like transistor gate voltages and channel conductivities, the same circuit structure can represent multiple weight levels, resolving the contradiction between information representation and structural complexity.
Data Source
AI summary
A neuromorphic device is provided. The neuromorphic device may include a pre-synaptic neuron, a synapse electrically connected with the pre-synaptic neuron through a row line, and a post-synaptic neuron electrically connected with the synapse through a column line. The post-synaptic neuron may include a post-neuron circuit and a post-neuron transfer function circuit electrically connected to the column line. The post-neuron transfer function circuit may include a first inverting circuit including at least one first pull-up transistor and at least two first pull-down transistors, the pull-down transistors being electrically connected with each other in parallel.