目录...................................................................................................................1...................................................................................................................2...................................................................................................2.......................................................................................2.......................................................................................3...................................................................................................4............................................................................5........................................................................................5................................................6.....................................................................................6............................................................8.........................................................................................9.............................................................................9........................................9.......................................................................12.......................................................................................12...............................................................................12.........................................................................12.........................................................13...............................................................16...................................................................................................16.................................................................................................................17.........................................................................................................17摘要【摘要】本文旨在探究相位幅度型复数值神经网络的可能性和必要性，并在数据集完成测试，得到肯定或否定的结果。本次设计将在充分了解bp（反向传播）神经网络和梯度下降法的相关知识的前提下，确定bp网络结构，反复讨论可行性后，确定设计方案。设计中主要完成将通常意义上的简单的实数值的神经网络转换成相位幅度型的复数值的神经网络，并在改写成相位幅度型的复数数据集上完成训练和测试，再对初始参数进行修改，以期最终达到百分之九十以上的正确率。此外，随着人工神经网络的研究逐步取得进展的同时，也开拓出更多新的天地，而复数值神经网络(CVNNs)则稳定地扩展了应用领域。【关键词】bp神经网络；相位幅度型复数；梯度下降法[Abstract]Thepurposeofthispaperistoexplorethepossibilityandnecessityofphase-magnitudecomplex-valuedneuralnetworks,andtotestthedatasetstoobtainpositiveornegativeresults.Thisdesignwilldeterminethebpnetworkstructureunderthepremiseoffullyunderstandingtherelevantknowledgeofthebp(back-propagation)neuralnetworkandthegradientdescentmethod.Afterrepeatedlydiscussingthefeasibility,thedesignplanisdetermined.Thedesignmainlycompletesthetransformationofasimplereal-valuedneuralnetworkintheusualsenseintoaphase-valuedcomplex-valuedneuralnetwork,andcompletesthetrainingandtestingonaphase-amplitudecomplexdataset,andthencarriesouttheinitialparameters.Revisedinordertoeventuallyachieveacorrectrateofmorethan90%.Inaddition,astheresearchonartificialneuralnetworksprogresses,moreandmorenewworldsareopenedup,andc...
VIP
