时间序列预测 | Matlab麻雀算法优化深度置信网络(SSA-DBN)时间序列预测

% pretrainDBN: pre-training the Deep Belief Nets (DBN) model by ContrastiveDivergence Learning %DBN的预训练：通过对比发散学习预训练深层信念网(DBN)模型。%%% dbn = pretrainDBN(dbn, V, opts)%对dbn进行预训练%%%Output parameters:%输出参数% dbn: the trained Deep Belief Nets (DBN) model%dbn：被训练的深度信念网络%%%Input parameters:%输入参数% dbn: the initial Deep Belief Nets (DBN) model%dbn：为最初的深度信念网络模型。% V: visible (input) variables, where # of row is number of data and # of col is # of visible (input) nodes%可见(输入)变量，第一行是数据的数量，而#是可见(输入)节点的#。% opts (optional): options%选择(可选):选项%% options (defualt value):%选项(默认值)%  opts.LayerNum: # of tarining RBMs counted from input layer (all layer)%opts.LayerNum：指训练中的受限玻尔兹曼机从输入层(所有层)的层数%  opts.MaxIter: Maxium iteration number (100)%opts.MaxIter马克西姆迭代数(100)%  opts.InitialMomentum: Initial momentum until InitialMomentumIter(0.5)%opts.InitialMomentum：初始动量，直到初始动量仪(0.5)%  opts.InitialMomentumIter: Iteration number for initial momentum(5)%opts.InitialMomentumIter：初始动量的迭代数(5)%  opts.FinalMomentum: Final momentum after InitialMomentumIter(0.9)%opts.FinalMomentum：初始动量之后的最终动量(0.9)%  opts.WeightCost: Weight cost (0.0002)%opts.WeightCost：权重(0.0002)%  opts.DropOutRate: List of Dropout rates for each layer(0)%opts.DropOutRate：每一层的退学率（0）%  opts.StepRatio: Learning step size (0.01)%opts.StepRatio：学习步长（0.01）%  opts.BatchSize: # of mini-batch data (# of all data)%opts.BatchSize：小批量数据#(所有数据的#)%  opts.Verbose: verbose or not (false)%opts.Verbose：详细的过程与否(错误)%  opts.SparseQ: q parameter of sparse learning (0)%opts.SparseQ：稀疏学习的q参数(0)%  opts.SparseLambda: lambda parameter (weight) of sparse learning(0)%opts.SparseLambda:λ参数(重量)的稀疏学习%%%Example:%举例% datanum = 1024;%数据数目% outputnum = 16;%输出数目% hiddennum = 8;%隐层数目% inputnum = 4;%输入数目% inputdata = rand(datanum, inputnum);%输入数据为一个datanum行，inputnum列的随机矩阵% outputdata = rand(datanum, outputnum);%输出矩阵为一个datanum行，outputnum列的随机矩阵% % dbn = randDBN([inputnum, hiddennum, outputnum]);%dbn为随机深度信念网络[输入数目，隐层数目，输出数目]% dbn = pretrainDBN( dbn, inputdata );dbn为预训练深度训练网络(dbn,输入数据)% dbn = SetLinearMapping( dbn, inputdata, outputdata );%调用SetLinearMapping函数% dbn = trainDBN( dbn, inputdata, outputdata );%调用trainDBN函数% % estimate = v2h( dbn, inputdata );%调用v2h函数%%%Reference:%参考%for details of the dropout%关于辍学的细节% Hinton et al, Improving neural networks by preventing co-adaptation of feature detectors, 2012.%Hinton等人，通过阻止功能探测器的协同适应，改善神经网络，2012年。%for details of the sparse learning%关于稀疏学习的细节。% Lee et al, Sparse deep belief net model for visual area V2, NIPS 2008.%Lee等人，稀疏的深层信仰网络模型的视觉区域V2, NIPS 2008。%%%Version: 20130821%版本：20130821%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Deep Neural Network:   %深度神经网络                      %%                                                          %% Copyright (C) 2013 Masayuki Tanaka. All rights reserved. %%                    mtanaka@ctrl.titech.ac.jp             %%    %版权(C) 2013年Masayuki Tanaka。保留所有权利。          %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%function dbn = pretrainDBN(dbn, V, opts)%pretrainDBN的调用格式LayerNum = numel( dbn.rbm );%LayerNum为dbn.rbm中元素的个数DropOutRate = zeros(LayerNum,1);%DropOutRate为X = V;%X为Vif( exist('opts' ) )%如果存在opts if( isfield(opts,'LayerNum') )%检查结构体opts是否包含由LayerNum指定的域， 如果包含，返回逻辑1; 如果opts不包含LayerNum域或者opts不是结构体类型的， 返回逻辑0。  LayerNum = opts.LayerNum;%LayerNum为opts.LayerNum end if( isfield(opts,'DropOutRate') )%检查结构体opts是否包含由DropOutRate指定的域， 如果包含，返回逻辑1; 如果opts不包含DropOutRate域或者opts不是结构体类型的， 返回逻辑0。  DropOutRate = opts.DropOutRate;%DropOutRate为opts.DropOutRate  if( numel( DropOutRate ) == 1 )%如果DropOutRate中的元素个数恒为1   DropOutRate = ones(LayerNum,1) * DropOutRate;%DropOutRate为LayerNum行，1列的全一矩阵乘以DropOutRate  end end else opts = [];%opts为空矩阵endfor i=1:LayerNum%i的取值范围为从1到LayerNum  opts.DropOutRate = DropOutRate(i);%opts.DropOutRate为DropOutRate(i)    dbn.rbm{i} = pretrainRBM(dbn.rbm{i}, X, opts);%dbn.rbm{1}调用pretrainRBM函数    X0 = X;%X0为X    X = v2h( dbn.rbm{i}, X0 );%X为调用v2h函数end