ANN Circuit Temperature Compensation via Crossbar Signal Summation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Artificial neural network (ANN) circuits with crossbar memristor circuits face performance deterioration due to varying temperature characteristics, leading to increased recognition errors in applications like image recognition, especially in non-cloud computing environments where low power consumption and high speed are crucial.
Innovation Solution
The ANN circuit design includes a crossbar circuit with multiple input and output bars intersecting via memristors, where the processing circuit calculates the sum of signals from multiple output bars to stabilize conductance values, reducing the impact of temperature variations and maintaining weight relations among memristors, and uses differential paired output bars to simulate excitatory and inhibitory synapses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a crossbar circuit with memristors is used for ANN calculations, then computation speed and energy efficiency are improved, but temperature variations cause conductance value changes leading to performance deterioration
Solution Approach 1:
The patent implements a feedback mechanism where the processing circuit monitors output signals from multiple output bars and adjusts input signals accordingly. The processing circuit calculates the sum of signals from multiple output bars and uses this information to compensate for temperature-induced conductance variations, thereby maintaining recognition accuracy while preserving the high-speed computation benefits of the crossbar circuit
Solution Approach 2:
The patent changes the operational parameters of the crossbar circuit by using multiple output bars and dynamically adjusting the number of bars involved in calculations based on temperature conditions. The processing circuit modifies signal processing parameters to compensate for conductance drift, allowing the system to maintain reliable performance across varying temperatures while utilizing the fast computation capability of the memristor crossbar
2Reliability
If multiple output bars are used to compensate for temperature effects, then recognition accuracy is improved, but circuit complexity increases
Solution Approach 1:
The patent makes the existing output bars serve multiple functions: they not only provide the primary calculation outputs but also serve as temperature compensation elements. By utilizing the same physical output bars for both computation and temperature effect mitigation, the patent avoids adding separate compensation circuits, thus improving reliability without proportionally increasing device complexity
Solution Approach 2:
The patent merges the temperature compensation function with the existing signal processing function by having the processing circuit simultaneously perform both the primary calculation and the temperature compensation using the same hardware resources. This consolidation allows the system to achieve improved recognition accuracy while minimizing the increase in circuit complexity
3Reliability
If conductance values are stabilized against temperature changes, then performance reliability is improved, but the ability to dynamically adjust weights is restricted
Solution Approach 1:
The patent segments the conductance value management into two distinct parts: (1) the base conductance values stored in memristors that remain relatively stable and represent the learned weights, and (2) the dynamic compensation factors applied by the processing circuit that adjust for temperature effects. This segmentation allows the system to maintain both performance consistency through stable base weights and adaptability through dynamic compensation adjustments
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 configuration effectively reduces the influence of conductance value variations with temperature changes, thereby suppressing performance deterioration and maintaining recognition accuracy across varying environmental temperatures.
Implementation Method 1
a plurality of memristors to give a weight to the signal to be transmitted, as a variable resistance memory
Implementation Method 2
the processing circuit calculates a sum of signals flowing into each of the output bars as signal processing in the layered neurons
Data Source
AI summary
An artificial neural network circuit includes a crossbar circuit, and a processing circuit. The crossbar circuit transmits a signal between layered neurons of an artificial neural network. The crossbar circuit includes input bars, output bars arranged intersecting the input bars, and memristors. The processing circuit calculates a sum of signals flowing into each of the output bars. The processing circuit calculates, as the sum of the signals, a sum of signals flowing into a plurality of separate output bars and conductance values of the corresponding memristors are set so as to cooperate to give a desired weight to the signal to be transmitted.