Font Size: a A A

Simulation And Analysis On STDP Of Electronic Synapse PCNE Based On Thermoelectric Effects

Posted on:2020-01-23Degree:MasterType:Thesis
Country:ChinaCandidate:Y LiFull Text:PDF
GTID:2428330590450364Subject:Software engineering
Abstract/Summary:PDF Full Text Request
With the deepening of the artificial intelligence,the increase of the amount of computation and the exposure of the Von Neumann bottleneck,the traditional computing model based on the binary and Von Neumann architecture is no longer efficient,the key to artificial intelligence is the construction of humanoid brain artificial neural network.The phase change nanometer element,which has both storage function and nano-integration,is one of the hot choices of electronic synapses in artificial neural networks.Spike-timing Dependent Plasticity is the important learning rule of synapse,when the size of the storage cell falls to the order of nanometer,the influence of thermoelectric effects on the cell can not be ignored,Therefore,the analysis on STDP of Electronic Synapse PCNE based on thermoelectric effects is of great significance.Based on the working principle,finite element method and thermoelectric effect of phase change nano-element,the physical and mathematical model of bottom-up PCNE three-dimensional structure is built in this paper.Under the environment of MATLAB,the electrical performance simulation,thermal performance simulation and the phase transition process simulation of PCNE are in progress by controlling the pulse parameters,element size and thermoelectric polarity.We can get the temperature,phase distribution and resistance changes of PCNE.The performance of PCNE electronic synapse and the realization of STDP learning rule under different thermoelectric effects are studied.In the three cases of positive and negative thermoelectric effects and non thermoelectric effects,The simulation results show that,The nonlinear output curves of electronic synapses LTD and LTP shift.The positive thermoelectric effect reduces the threshold voltage of PCNE in LTD process by 11.7%.Taking LTD process as an example,changing the size parameters of PCNE structure,including the size of heating pole Ti N and the thickness of phase change material GST,the simulation results show that the smaller the size of Ti N and the thickness of GST,the weaker the influence of thermoelectric effect on PCNE element.Based on STDP learning rule,the pre-synaptic and post-synaptic neuron pulse sequences are designed,and the STDP learning curves under positive and negativethermoelectric effects and non-thermoelectric effects are drawn.It was found that the amplitude of pre-synaptic neurons was lower under the positive thermal effect.Simultaneously,the smaller the size of Ti N and the thickness of GST,the weaker the influence of thermal effect on the amplitude range of pre-synaptic neurons.
Keywords/Search Tags:Artificial neural networks, Phase change nanometer element(PCNE), Electronic synapses, STDP learning rules, Finite element method, thermoelectric effects
PDF Full Text Request
Related items