Matlab Neural Network

download Matlab Neural Network

of 9

Transcript of Matlab Neural Network

  • 8/9/2019 Matlab Neural Network

    1/9

     

    January 2-, 2015, MATLAB Tutorial for ECE 614

    Regression Using eural et!or"

    1- Using eural et!or" #U$ %nnstart&

  • 8/9/2019 Matlab Neural Network

    2/9

     

    2- Using Co''an(-Line fun)tions

    % Solve an Input-Output Fitting problem with a Neural Network % Script generated by NFTOOL % % This script assumes these variables are defined % % houseInputs - input data! % houseTargets - target data! 

    load house"datasetinputs # houseInputs$targets # houseTargets$

    % reate a Fitting Network hiddenLayerSi&e # '($net # fitnet)hiddenLayerSi&e*$

    % Set up +ivision of +ata for Training, alidation, Testing net!divide.aram!train/atio # 0(1'(($net!divide.aram!val/atio # '21'(($net!divide.aram!test/atio # '21'(($

    % Train the Network 3net,tr4 # train)net,inputs,targets*$

    % Test the Network outputs # net)inputs*$errors # gsubtract)outputs,targets*$

    performance # perform)net,targets,outputs*

    % iew the Network view)net*

    % .lots % 5ncomment these lines to enable various plots! % figure, plotperform)tr* % figure, plottrainstate)tr* % figure, plotfit)targets,outputs* % figure, plotregression)targets,outputs* 

    % figure, ploterrhist)errors*

  • 8/9/2019 Matlab Neural Network

    3/9

     

    Classifi)ation Using eural et!or"

    1- Using eural et!or" #U$ %nnstart&

  • 8/9/2019 Matlab Neural Network

    4/9

     

    2- Using Co''an(-Line fun)tions

    % Solve a .attern /ecognition .roblem with a Neural Network % Script generated by N./TOOL % % This script assumes these variables are defined % % cancerInputs - input data! % cancerTargets - target data! 

    inputs # cancerInputs$targets # cancerTargets$

    % reate a .attern /ecognition Network hiddenLayerSi&e # '($net # patternnet)hiddenLayerSi&e*$

    % Set up +ivision of +ata for Training, alidation, Testing net!divide.aram!train/atio # 0(1'(($net!divide.aram!val/atio # '21'(($net!divide.aram!test/atio # '21'(($

    % Train the Network 3net,tr4 # train)net,inputs,targets*$

    % Test the Network outputs # net)inputs*$errors # gsubtract)targets,outputs*$performance # perform)net,targets,outputs*

    % iew the Network view)net*

    % .lots % 5ncomment these lines to enable various plots! % figure, plotperform)tr* % figure, plottrainstate)tr* % figure, plotconfusion)targets,outputs* % figure, ploterrhist)errors*

  • 8/9/2019 Matlab Neural Network

    5/9

     

    Clustering Using eural et!or" %*elf-+rganiing

    Mas&

    1- Using eural et!or" #U$ %nnstart&

    2- Using Co''an(-Line fun)tions

  • 8/9/2019 Matlab Neural Network

    6/9

     

    % Solve a lustering .roblem with a Self-Organi&ing 6ap % Script generated by NTOOL % % This script assumes these variables are defined % % simpleclusterInputs - input data! 

    load iris"dataset inputs # irisInputs$

    % reate a Self-Organi&ing 6ap dimension' # '($dimension7 # '($net # selforgmap)3dimension' dimension74*$

    % Train the Network 3net,tr4 # train)net,inputs*$

    % Test the Network outputs # net)inputs*$

    % iew the Network view)net*

    % .lots % 5ncomment these lines to enable various plots! % figure, plotsomtop)net* % figure, plotsomnc)net* 

    % figure, plotsomnd)net* % figure, plotsomplanes)net* % figure, plotsomhits)net,inputs* % figure, plotsompos)net,inputs*

  • 8/9/2019 Matlab Neural Network

    7/9

     

    E.a'les fro' t/e Te.t

    1-igit ete)tion

    %% data as in e!g! 8!'9, p! 228

    digitsInput # 3!!!( ( ' ( ( !!! here goes the '( ( ' ( ( !!!( ( ' ( ( !!!( ( ' ( ( !!!

    ( ( ' ( (( ' ' ' ( !!! here goes the 7( ( ( ' ( !!!( ( ' ( ( !!!( ' ( ( ( !!!' ' ' ' '( ' ' ' ( !!!here goes the :( ( ( ' ( !!!( ( ' ' ( !!!( ( ( ' ( !!!( ' ' ' (( ' ( ( ' !!! ;( ' ( ( ' !!!( ' ' ' ' !!!( ( ( ( ' !!!( ( ( ( '( ' ' ' ( !!! 2( ' ( ( ( !!!( ' ' ' ( !!!( ( ( ' ( !!!( ' ' ' (4$

    digitsInput#digitsInput

  • 8/9/2019 Matlab Neural Network

    8/9

     

    %% plot the digitsfigure$for d#'2

    subplot)',2,d*$imshow)reshape)digitsInput),d*,2,2*#(!8?sin)pi?=*$

    %plot)=,>*

    net#newfit)=,>,2*$net!divideFcn#,

  • 8/9/2019 Matlab Neural Network

    9/9

     

    in#=Test),i*$myOut),i*#O@?tansig)B@?inGBC*GOC$% H input neuronsH% H output activation, no output sigmoid

    end%postprocess the outputmyOut#mapminma)