ArtificialNeuralNetworks人工神经网络.pptxVIP

ArtificialNeuralNetworks人工神经网络会计学Supervised ANNsExamplesApplicationsFurther topicsUnsupervised ANNsExamplesApplicationsFurther topicsIIIII第1页/共69页Table of ContentsIntroduction to ANNsTaxonomyFeaturesLearningApplicationsIArtificial Neural Networks - I第2页/共69页Contents - IIntroduction to ANNsProcessing elements (neurons)ArchitectureFunctional Taxonomy of ANNsStructural Taxonomy of ANNsFeaturesLearning ParadigmsApplicationsArtificial Neural Networks - I billion synapses in human brainChemical transmission and modulation of signalsInhibitory synapsesExcitatory synapses第3页/共69页The Biological Neuron 10 billion neurons in human brainSummation of input stimuliSpatial (signals)Temporal (pulses)Threshold over composed inputsConstant firing strengthArtificial Neural Networks - I第4页/共69页Biological Neural Networks10,000 synapses per neuronComputational power = connectivityPlasticity new connections (?)strength of connections modified Artificial Neural Networks - I第5页/共69页Neural DynamicsAction potentialAction potential ≈ 100mVActivation threshold ≈ 20-30mVRest potential ≈ -65mVSpike time ≈ 1-2msRefractory time ≈ 10-20msRefractory timeArtificial Neural Networks - I第6页/共69页神经网络的复杂性神经网路的复杂多样，不仅在于神经元和突触的数量大、组合方式复杂和联系广泛，还在于突触传递的机制复杂。现在已经发现和阐明的突触传递机制有：突触后兴奋，突触后抑制，突触前抑制，突触前兴奋，以及“远程”抑制等等。在突触传递机制中，释放神经递质是实现突触传递机能的中心环节，而不同的神经递质有着不同的作用性质和特点Artificial Neural Networks - I第7页/共69页神经网络的研究神经系统活动，不论是感觉、运动，还是脑的高级功能（如学习、记忆、情绪等）都有整体上的表现，面对这种表现的神经基础和机理的分析不可避免地会涉及各种层次。这些不同层次的研究互相启示，互相推动。在低层次（细胞、分子水平）上的工作为较高层次的观察提供分析的基础，而较高层次的观察又有助于引导低层次工作的方向和体现其功能意义。既有物理的、化学的、生理的、心理的分门别类研究，又有综合研究。Artificial Neural Networks - I第8页/共69页The Artificial Neuronx1(t)Stimuluswi1x2(t)wi2wi3x3(t)yi(t)wi4Responsex4(t)wi5x5(t)Neuron iurest = resting potentialxj(t) = output of neuron j at time twij = connection strength between neuron i and neuron ju(t) = total stimulus at time tArtificial Neural Networks - I第9页/共69页Artificial Neural ModelsMcCulloch Pitts-type Neurons (static)Digital neurons: activation state interpretation (snapsh