FPGA-based stdp synaptic plasticity experiment platform under feed-forward neural network

Abstract

The invention discloses an FPGA (Field Programmable Gate Array)-based STDP (Spike Timing-dependent Plasticity) synaptic plasticity experimental platform under feedforward neural network. The experimental platform comprises an FPGA development board and an upper computer which are connected with each other, wherein the FPGA is taken as a lower computer and is provided with a human-machine operation interface programmed by C++ to perform synaptic strength connection change waveform and dynamical characteristic observation and parameter setting; the FPGA is used for implementing a mathematic model, and applying an external stimulate signal to a multilayer feedforward neural network and a synaptic connection model; the upper computer is applied to dynamics analysis such as parameter adjustment, synaptic change waveform observation, synaptic adaptive change and the like. The FPGA-based STDP synaptic plasticity experimental platform has the effects that on the basis of a high-speed operation FPGA neural synaptic plasticity computation platform, hardware modeling relative to a synaptic plasticity part among phenotype nerve cells is realized through a non-animal experiment of biological neural synaptic connection, the learning process can be stabilized effectively, and the consistence with real neuronal synapse plastic connection is achieved.

Description

technical field [0001] The invention relates to biomedical engineering technology, in particular to an FPGA-based STDP synaptic plasticity experiment platform under a feedforward neural network. Background technique [0002] In the biological nervous system, a large number of neurons are interconnected through synapses to form neural circuits. Synapses are important parts of neuron information transmission, and the communication between neurons also depends on synapses as a medium. In a neuronal network, the different connections of synapses also affect the function of its neural network. Excitation timing-dependent synaptic plasticity is a type of synaptic plasticity driven by precise timing differences between presynaptic and postsynaptic action potentials. Therefore, STDP-based learning rules are suitable for learning some time-related neural phenomena, such as action potential-time synchronization. The traditional weight-independent STDP learning rule creates an unstab...

AI Technical Summary

Problems solved by technology

[0005] The existing technology is still in the basic stage, so there are still the following disadvantages: there is no FPGA-based dedicated STDP synaptic plasticity experiment platform with complete functions; the hardware simulation synapse calculation model using FPGA is relatively simple in structure and low in accuracy; The computer interface is not yet perfect, and real-time control operations and data analysis cannot be performed. Therefore, it is difficult to operate and analyze the dynamic characteristics of the synaptic weight changes of FPGA hardware neurons.

Images