Threshold-Switching Analog-Stochastic Converter for Parallel Weight Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional weight updating methods in neuromorphic technologies, which involve programming synaptic elements, are inefficient as they require external calculation and transmission of weight change values, leading to increased time requirements with larger synaptic arrays, especially when using a column-by-column or row-by-row format.
Innovation Solution
An analog-stochastic converter is introduced, utilizing a threshold switching element and a probability conversion circuit to convert analog voltage signals into pulse signals with corresponding probabilities, enabling simultaneous application of probability signals to all row and column lines for fully-parallel weight updating, thereby reducing the time required for weight updates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional column-by-column or row-by-row weight updating method is used, then weight updating can be performed systematically, but the time required for updating increases significantly when the size of synaptic array is increased
Solution Approach 1:
The patent divides the weight updating process into two independent stages: (1) converting analog input signals to probability signals using threshold switching elements in each column independently, and (2) applying these probability signals to select synapses for weight updates. This segmentation allows parallel processing across multiple columns simultaneously, breaking the sequential bottleneck of conventional methods.
Solution Approach 2:
The patent replaces the conventional mechanical/sequential weight updating mechanism with a stochastic/probabilistic mechanism. By using threshold switching elements that convert analog voltages to probability signals, the system enables parallel stochastic selection of synapses across the entire array, substituting the sequential row-by-row or column-by-column approach with simultaneous probabilistic updates throughout the array.
2Ease of operation
If external calculation and transmission of weight change values is used, then weight updates can be performed with external control, but the process becomes inefficient and time-consuming
Solution Approach 1:
The patent implements self-service by enabling the synaptic array to perform its own weight updates through intrinsic stochastic mechanisms. Each threshold switching element automatically converts input voltages to probability signals that directly control synapse selection, eliminating the need for external calculation and transmission of weight change values. The system uses its own internal resources (threshold switching elements and probability conversion) to accomplish weight updates autonomously.
Solution Approach 2:
The patent introduces threshold switching elements as intermediary components between the input signal source and the synaptic weights. These intermediaries convert analog input voltages into probability signals that stochastically control synapse activation, serving as a bridge that translates external inputs into internal weight updates without requiring external calculation systems.
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
This approach allows for faster weight updating operations by converting analog signals into probability signals, which are applied in parallel, significantly reducing the time needed for updating synaptic weights compared to conventional methods.
Implementation Method 1
a threshold switching element configured to receive a pulse signal corresponding to an analog signal and be turned on according to the input pulse signal
Data Source
AI summary
There is provided an analog-stochastic converter for converting an analog voltage signal into a pulse signal having a corresponding probability. The analog-stochastic converter is implemented using a threshold switching element and a simple logic circuit, thereby reducing a size of the analog-stochastic converter and enabling a low power operation thereof. In addition, in order to update a weight, instead of an analog signal, a probability signal is applied using the above-described analog-stochastic converter, thereby updating a weight in a fully-parallel manner in a synaptic element array having an intersection structure. Accordingly, it is possible to shorten a time for weight update.