Synapse Circuit STDP Weight Writing With Self-Timed Spike Pulses
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for storing synaptic coupling weights in synapse circuits based on spike timing dependent synaptic plasticity (STDP) are complex, leading to increased power consumption and circuit scale, necessitating a simplified driving method.
Innovation Solution
A synapse circuit with a weight storage unit that stores synaptic coupling weights based on pulse time differences and order, using specific time window pulses and write pulses to simplify the writing process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If learning based on STDP is performed with complex control for writing synaptic coupling weights according to pulse order and time difference, then the synaptic coupling weight can be stored accurately, but the control complexity increases and power consumption increases
Solution Approach 1:
The synapse circuit automatically determines write timing by detecting the relative timing of pre-spike and post-spike pulses itself, without requiring external control signals. The circuit uses its own input pulses to generate the write operation, eliminating the need for complex external control mechanisms while maintaining accurate STDP-based weight storage
Solution Approach 2:
The synapse circuit prepares the write operation in advance by continuously monitoring the timing relationship between pre-spike and post-spike pulses. When the appropriate timing condition is met (post-spike precedes pre-spike), the circuit is ready to immediately perform the write operation without requiring complex real-time control decisions
2Manufacturing precision
If learning based on STDP is performed with complex control for writing synaptic coupling weights, then the synaptic coupling weight can be stored accurately, but power consumption increases
Solution Approach 1:
The synapse circuit uses periodic spike pulses from pre-neuron and post-neuron as the basis for write operations. Instead of continuous control signaling, the circuit leverages the natural periodic spiking activity to trigger write operations only when the specific timing condition (post-spike preceding pre-spike) occurs, significantly reducing power consumption while maintaining accurate weight storage
Solution Approach 2:
The circuit uses its own input spike pulses to directly control the write operation timing, eliminating the need for separate power-consuming control signal generation. The write operation is automatically triggered by the timing relationship between the input pulses themselves, reducing overall power consumption
3Manufacturing precision
If complex control is used for writing synaptic coupling weights based on pulse order and time difference, then the synaptic coupling weight can be stored accurately, but the circuit scale increases
Solution Approach 1:
The synapse circuit combines the functions of pulse timing detection, write timing determination, and weight storage into a single integrated circuit. By merging these functions rather than implementing them as separate control modules, the circuit achieves accurate STDP-based weight storage without increasing overall circuit scale
Solution Approach 2:
The circuit uses its own input pulses to control its own write operations, eliminating the need for external control circuitry. This self-controlling mechanism reduces the circuit scale by removing redundant control paths and components that would otherwise be needed to manage the write timing
Data Source
AI summary
Provided is a simplified driving method of a synapse circuit. In a case where a first pre-spike pulse precedes a first post-spike pulse, a second pre-spike pulse from an input circuit is used as a time window that allows writing of a coupling weight, and the first post-spike pulse from a neuron circuit is used as a write pulse for controlling a write timing of the coupling weight. In a case where the first post-spike pulse precedes the first pre-spike pulse, a second post-spike pulse from the neuron circuit is used as the time window, and the first pre-spike pulse from the input circuit is used as the write pulse. The second pre-spike pulse and the second post-spike pulse are output in synchronization with the first pre-spike pulse and the first post-spike pulse, respectively.