I'm running an example that I got from a Webinar.
this is the code:
%% Fine Tuning A Deep Neural Network
clear; clc;close all;
imagenet_cnn = load('imagenet-cnn');
net = imagenet_cnn.convnet;
net.Layers
%% Perform net surgery
layers = net.Layers(1:end-3);
layers(end+1) = fullyConnectedLayer(12, 'Name', 'fc8_2')
layers(end+1) = softmaxLayer('Name','prob_2');
layers(end+1) = classificationLayer('Name','classificationLayer_2')
%% Setup learning rates for fine-tuning
% fc 8 - bump up learning rate for last layers
layers(end-2).WeightLearnRateFactor = 100;
layers(end-2).WeightL2Factor = 1;
layers(end-2).BiasLearnRateFactor = 20;
layers(end-2).BiasL2Factor = 0;
%% Load Image Data
rootFolder = fullfile('E:\Universidad\Tesis\Matlab', 'TesisDataBase');
categories = {'Avion','Banana','Carro','Gato', 'Mango','Perro','Sandia','Tijeras','Silla','Mouse','Calculadora','Arbol'};
imds = imageDatastore(fullfile(rootFolder, categories), 'LabelSource', 'foldernames');
tbl = countEachLabel(imds);
%% Equalize number of images of each class in training set
minSetCount = min(tbl{:,2}); % determine the smallest amount of images in a category
% Use splitEachLabel method to trim the set.
imds = splitEachLabel(imds, minSetCount);
% Notice that each set now has exactly the same number of images.
countEachLabel(imds)
[trainingDS, testDS] = splitEachLabel(imds, 0.7,'randomize');
% Convert labels to categoricals
trainingDS.Labels = categorical(trainingDS.Labels);
trainingDS.ReadFcn = #readFunctionTrain;
%% Setup test data for validation
testDS.Labels = categorical(testDS.Labels);
testDS.ReadFcn = #readFunctionValidation;
%% Fine-tune the Network
miniBatchSize = 32; % lower this if your GPU runs out of memory.
numImages = numel(trainingDS.Files);
numIterationsPerEpoch = 250;
maxEpochs = 62;
lr = 0.01;
opts = trainingOptions('sgdm', ...
'InitialLearnRate', lr,...
'LearnRateSchedule', 'none',...
'L2Regularization', 0.0005, ...
'MaxEpochs', maxEpochs, ...
'MiniBatchSize', miniBatchSize);
net = trainNetwork(trainingDS, layers, opts);
As you can see this code , uses the well known AlexNet as a first start, then the last 3 layers are deleted ,in order to put 3 new layers with the number of neurons necessary for the new task.
the read func for test and training are the same here you have one of them:
function Iout = readFunctionTrain(filename)
% Resize the flowers images to the size required by the network.
I = imread(filename);
% Some images may be grayscale. Replicate the image 3 times to
% create an RGB image.
if ismatrix(I)
I = cat(3,I,I,I);
end
% Resize the image as required for the CNN.
Iout = imresize(I, [227 227]);
this code runs well at the webinar, they use it to classify cars and subs that pass thru the matworks door.
The problem is that the new net is not learning when I try it with my own images,I have a data set of 12 categories each one with 1000 images more or less, all this images where downloaded from ImageNET.
the net does not increase its Mini batch accuracy, actually some times it does but very slow.
I also did the tutorial of this page
Matlab Deep Learning ToolBox
and it worked good with my images. So , I don't understand what is wrong with my fine-tuning. Thanks.
If you have R2016a and a GeForce GTX1080 or other Pascal GPU, then see this tech support answer and this bug report.
Your learning rate for the pre-trained section of the network (0.01) looks very high for a fine tuning workflow. Also, your LR of 1.0 is quite high for the randomly initialized classification head.
What happens if you set the learning rate of the pre-trained section to 0 and train only the randomly initialized head of the network? What happens if you just use a low learning rate and train end to end (say 1e-5)?
It would be useful to see the training-progress plot, but I think its possible you're not converging due to your learning rate settings.
Related
I want to ask your help in EEG data classification.
I am a graduate student trying to analyze EEG data.
Now I am struggling with classifying ERP speller (P300) with SWLDA using Matlab
Maybe there is something wrong in my code.
I have read several articles, but they did not cover much details.
My data size is described as below.
size(target) = [300 1856]
size(nontarget) = [998 1856]
row indicates the number of trials, column indicates spanned feature
(I stretched data [64 29] (for visual representation I did not select ROI)
I used stepwisefit function in Matlab to classify target vs non-target
Code is attached below.
ingredients = [targets; nontargets];
heat = [class_targets; class_nontargets]; % target: 1, non-target: -1
randomized_set = shuffle([ingredients heat]);
for k=1:10 % 10-fold cross validation
parition_factor = ceil(size(randomized_set,1) / 10);
cv_test_idx = (k-1)*parition_factor + 1:min(k * parition_factor, size(randomized_set,1));
total_idx = 1:size(randomized_set,1);
cv_train_idx = total_idx(~ismember(total_idx, cv_test_idx));
ingredients = randomized_set(cv_train_idx, 1:end-1);
heat = randomized_set(cv_train_idx, end);
[W,SE,PVAL,INMODEL,STATS,NEXTSTEP,HISTORY]= stepwisefit(ingredients, heat, 'penter', .1);
valid_id = find(INMODEL==1);
v_weights = W(valid_id)';
t_ingredients = randomized_set(cv_test_idx, 1:end-1);
t_heat = randomized_set(cv_test_idx, end); % true labels for test set
v_features = t_ingredients(:, valid_id);
v_weights = repmat(v_weights, size(v_features, 1), 1);
predictor = sum(v_weights .* v_features, 2);
m_result = predictor > 0; % class A: +1, B: 0
t_heat(t_heat==-1) = 0;
acc(k) = sum(m_result==t_heat) / length(m_result);
end
p.s. my code is currently very inefficient and might be bad..
In my assumption, stepwisefit calculates significant coefficients every steps, and valid column would be remained.
Even though it's not LDA, but for binary classification, LDA and linear regression are not different.
However, results were almost random chance.. (for other binary data on the internet, it worked..)
I think I made something wrong, and your help can correct me.
I will appreciate any suggestion and tips to implement classifier for ERP speller.
Or any idea for implementing SWLDA in Matlab code?
The name SWLDA is only used in the context of Brain Computer Interfaces, but I bet it has another name in a more general context.
If you track the recipe of SWLDA you will end up in Krusienski 2006 papers ("A comparison..." and "Toward enhanced P300..") and from there the book where stepwise logarithmic regression is explained: "Draper Smith, Applied Regression Analysis, 1981". However, as far as I am aware of, no paper gives actually the complete recipe on how to implement it (and their details and secrets).
My approach was using stepwiseglm:
H=predictors;
TH=variables;
lbs=labels % (1,2)
if (stepwiseflag)
mdl = stepwiseglm(H', lbs'-1,'constant','upper','linear','distr','binomial');
if (mdl.NumEstimatedCoefficients>1)
inmodel = [];
for i=2:mdl.NumEstimatedCoefficients
inmodel = [inmodel str2num(mdl.CoefficientNames{i}(2:end))];
end
H = H(inmodel,:);
TH = TH(inmodel,:);
end
end
lbls = classify(TH',H',lbs','linear');
You can also use a k-fold cross validaton approach using matlab cvpartition.
c = cvpartition(lbs,'k',10);
opts = statset('display','iter');
fun = #(XT,yT,Xt,yt)...
(sum(~strcmp(yt,classify(Xt,XT,yT,'linear'))));
I'm new in machine learning (and to stackoverflow as well) and i want to make some classification tasks. I performed two group classifications on my data set (field of speech acoustics) with LIBSVM and Matlab's Pattern Recignition Tool from the Neural network toolbox to create a simple network with one hidden layer. In the hope of higher classification results i want to try Deep Neural Networks, and i found this code: http://www.mathworks.com/matlabcentral/fileexchange/42853-deep-neural-network
I have some difficulty understanding it.
My data is constructed of 127 samples of 19 parameters, so my input number is 19. I want to classify them in two groups: 0 and 1, so my output number is 1. The values in my data set are normalized between 0 and 1.
My code is the following:
clear all
clc
addpath('..');
load('data.mat')
inputdata = inputs;
outputdata = outputs;
datanum = 127;
outputnum = 1;
hiddennum = 3;
inputnum = 19;
% rbm = randRBM(inputnum, outputnum);
% rbm = pretrainRBM( rbm, inputdata );
dbn = randDBN([inputnum, hiddennum, outputnum]);
dbn = pretrainDBN( dbn, inputdata );
dbn = SetLinearMapping( dbn, inputdata, outputdata );
dbn = trainDBN( dbn, inputdata, outputdata );
estimate = v2h( dbn, inputdata )
[rmse AveErrNum] = CalcRmse(dbn, inputdata, outputdata)
The code runs. The rmse is 0.4183, the AveErrNum is 0.1969. What i need is the classification accuracy between my targets (stored in outputdata) and the networks predictions (Accuracy = data classified correctly / all data).
Where do i find the networks predictions after binarization?
Do I use the right type of network for my classification?
Don't I need to divide my data into Training, Validation and Testing samples (like in the case of a simple neural network with one hidden layer)?
Thanks in advance for any help!
I am trying to implement Bag of Words in opencv and has come with the implementation below. I am using Caltech 101 database. However, since its my first time and not being familiar, I have planned to used two image sets from the database, the chair image set and the soccer ball image set. I have coded for the svm using this.
Everything went allright, except when I call classifier.predict(descriptor) , I do not get the label vale as intended. I always get a0 instead of '1', irrespective of my test image. The number of images in the chair dataset is 10 and in the soccer ball dataset is 10. I labelled chair as 0 and soccer ball as 1 . The links represent the samples of each categories, the top 10 is of chairs, the bottom 10 is of soccer balls
function hello
clear all; close all; clc;
detector = cv.FeatureDetector('SURF');
extractor = cv.DescriptorExtractor('SURF');
links = {
'http://i.imgur.com/48nMezh.jpg'
'http://i.imgur.com/RrZ1i52.jpg'
'http://i.imgur.com/ZI0N3vr.jpg'
'http://i.imgur.com/b6lY0bJ.jpg'
'http://i.imgur.com/Vs4TYPm.jpg'
'http://i.imgur.com/GtcwRWY.jpg'
'http://i.imgur.com/BGW1rqS.jpg'
'http://i.imgur.com/jI9UFn8.jpg'
'http://i.imgur.com/W1afQ2O.jpg'
'http://i.imgur.com/PyX3adM.jpg'
'http://i.imgur.com/U2g4kW5.jpg'
'http://i.imgur.com/M8ZMBJ4.jpg'
'http://i.imgur.com/CinqIWI.jpg'
'http://i.imgur.com/QtgsblB.jpg'
'http://i.imgur.com/SZX13Im.jpg'
'http://i.imgur.com/7zVErXU.jpg'
'http://i.imgur.com/uUMGw9i.jpg'
'http://i.imgur.com/qYSkqEg.jpg'
'http://i.imgur.com/sAj3pib.jpg'
'http://i.imgur.com/DMPsKfo.jpg'
};
N = numel(links);
trainer = cv.BOWKMeansTrainer(100);
train = struct('val',repmat({' '},N,1),'img',cell(N,1), 'pts',cell(N,1), 'feat',cell(N,1));
for i=1:N
train(i).val = links{i};
train(i).img = imread(links{i});
if ndims(train(i).img > 2)
train(i).img = rgb2gray(train(i).img);
end;
train(i).pts = detector.detect(train(i).img);
train(i).feat = extractor.compute(train(i).img,train(i).pts);
end;
for i=1:N
trainer.add(train(i).feat);
end;
dictionary = trainer.cluster();
extractor = cv.BOWImgDescriptorExtractor('SURF','BruteForce');
extractor.setVocabulary(dictionary);
for i=1:N
desc(i,:) = extractor.compute(train(i).img,train(i).pts);
end;
a = zeros(1,10)';
b = ones(1,10)';
labels = [a;b];
classifier = cv.SVM;
classifier.train(desc,labels);
test_im =rgb2gray(imread('D:\ball1.jpg'));
test_pts = detector.detect(test_im);
test_feat = extractor.compute(test_im,test_pts);
val = classifier.predict(test_feat);
disp('Value is: ')
disp(val)
end
These are my test samples:
Soccer Ball
(source: timeslive.co.za)
Chair
Searching through this site I think that my algorithm is okay, even though I am not quite confident about it. If anybody can help me in finding the bug, it will be appreciable.
Following Amro's code , this was my result:
Distribution of classes:
Value Count Percent
1 62 49.21%
2 64 50.79%
Number of training instances = 61
Number of testing instances = 65
Number of keypoints detected = 38845
Codebook size = 100
SVM model parameters:
svm_type: 'C_SVC'
kernel_type: 'RBF'
degree: 0
gamma: 0.5063
coef0: 0
C: 62.5000
nu: 0
p: 0
class_weights: 0
term_crit: [1x1 struct]
Confusion matrix:
ans =
29 1
1 34
Accuracy = 96.92 %
Your logic looks fine to me.
Now I guess you'll have to tweak the various parameters if you want to improve the classification accuracy. This includes the clustering algorithm parameters (such as the vocabulary size, clusters initialization, termination criteria, etc..), the SVM parameters (kernel type, the C coefficient, ..), the local features algorithm used (SIFT, SURF, ..).
Ideally, whenever you want to perform parameter selection, you ought to use cross-validation. Some methods already have such mechanism embedded (CvSVM::train_auto for instance), but for the most part you'll have to do this manually...
Finally you should follow general machine learning guidelines; see the whole bias-variance tradeoff dilemma. The online Coursera ML class discusses this topic in detail in week 6, and explains how to perform error analysis and use learning curves to decide what to try next (do we need to add more instances, increase model complexity, and so on..).
With that said, I wrote my own version of the code. You might wanna compare it with your code:
% dataset of images
% I previously saved them as: chair1.jpg, ..., ball1.jpg, ball2.jpg, ...
d = [
dir(fullfile('images','chair*.jpg')) ;
dir(fullfile('images','ball*.jpg'))
];
% local-features algorithm used
detector = cv.FeatureDetector('SURF');
extractor = cv.DescriptorExtractor('SURF');
% extract local features from images
t = struct();
for i=1:numel(d)
% load image as grayscale
img = imread(fullfile('images', d(i).name));
if ~ismatrix(img), img = rgb2gray(img); end
% extract local features
pts = detector.detect(img);
feat = extractor.compute(img, pts);
% store along with class label
t(i).img = img;
t(i).class = find(strncmp(d(i).name,{'chair','ball'},4));
t(i).pts = pts;
t(i).feat = feat;
end
% split into training/testing sets
% (a better way would be to use cvpartition from Statistics toolbox)
disp('Distribution of classes:')
tabulate([t.class])
tTrain = t([1:7 11:17]);
tTest = t([8:10 18:20]);
fprintf('Number of training instances = %d\n', numel(tTrain));
fprintf('Number of testing instances = %d\n', numel(tTest));
% build visual vocabulary (by clustering training descriptors)
K = 100;
bowTrainer = cv.BOWKMeansTrainer(K, 'Attempts',5, 'Initialization','PP');
clust = bowTrainer.cluster(vertcat(tTrain.feat));
fprintf('Number of keypoints detected = %d\n', numel([tTrain.pts]));
fprintf('Codebook size = %d\n', K);
% compute histograms of visual words for each training image
bowExtractor = cv.BOWImgDescriptorExtractor('SURF', 'BruteForce');
bowExtractor.setVocabulary(clust);
M = zeros(numel(tTrain), K);
for i=1:numel(tTrain)
M(i,:) = bowExtractor.compute(tTrain(i).img, tTrain(i).pts);
end
labels = vertcat(tTrain.class);
% train an SVM model (perform paramter selection using cross-validation)
svm = cv.SVM();
svm.train_auto(M, labels, 'SvmType','C_SVC', 'KernelType','RBF');
disp('SVM model parameters:'); disp(svm.Params)
% evaluate classifier using testing images
actual = vertcat(tTest.class);
pred = zeros(size(actual));
for i=1:numel(tTest)
descs = bowExtractor.compute(tTest(i).img, tTest(i).pts);
pred(i) = svm.predict(descs);
end
% report performance
disp('Confusion matrix:')
confusionmat(actual, pred)
fprintf('Accuracy = %.2f %%\n', 100*nnz(pred==actual)./numel(pred));
Here are the output:
Distribution of classes:
Value Count Percent
1 10 50.00%
2 10 50.00%
Number of training instances = 14
Number of testing instances = 6
Number of keypoints detected = 6300
Codebook size = 100
SVM model parameters:
svm_type: 'C_SVC'
kernel_type: 'RBF'
degree: 0
gamma: 0.5063
coef0: 0
C: 312.5000
nu: 0
p: 0
class_weights: []
term_crit: [1x1 struct]
Confusion matrix:
ans =
3 0
1 2
Accuracy = 83.33 %
So the classifier correctly labels 5 out of 6 images from the test set, which is not bad for a start :) Obviously you'll get different results each time you run the code due to the inherent randomness of the clustering step.
What is the number of images you are using to build your dictionary i.e. what is N? From your code, it seems that you are only using a 10 images (those listed in links). I hope this list is truncated down for this post else that would be too few. Typically you need in the order of 1000 or much more images to build the dictionary and the images need not be restricted to only these 2 classes that you are classifying. Otherwise, with only 10 images and 100 clusters your dictionary is likely to be messed up.
Also, you might want to use SIFT as a first choice as it tends to perform better than the other descriptors.
Lastly, you can also debug by checking the detected keypoints. You can get OpenCV to draw the keypoints. Sometimes your keypoint detector parameters are not set properly, resulting in too few keypoints getting detected, which in turn gives poor feature vectors.
To understand more about the BOW algorithm, you can take a look at these posts here and here. The second post has a link to a free pdf for an O'Reilley book on computer vision using python. The BOW model (and other useful stuff) is described in more details inside that book.
Hope this helps.
I'm trying to set up a custom neural network, but when I train it, it doesn't train : the training process makes 0 iterations! I don't get any errors though, just 0 iterations, and I have no idea why. (The architecture might seem odd to you, it is supposed to be a custom PNN. But before we can even discuss if it makes sense or not, I would like to be able to train it...)
Here is the code
net = network;
net.trainFcn = 'trainlm';
net.performFcn = 'mse';
net.numInputs = 1;
net.numLayers = (2*nbclasses)+1; % (one pattern layer + one summation layer per class) + competition layer
net.inputConnect(1:nbclasses,:) = 1; % connects the input to all pattern layers
for i = 1:nbclasses % Connect the pattern layers to their corresponding summation layers
net.layerConnect(i+nbclasses,i) = 1;
net.layers{i}.size = size(tr_feature,1);
net.layers{i}.transferFcn = 'radbas';
end
for i = (nbclasses+1):(nbclasses*2) % Connect all summation layers to the competition layer
net.layers{i}.size = 1;
net.layerConnect(net.numLayers,i) = 1;
end
net.layers{net.numLayers}.transferFcn = 'compet';
net.outputConnect(1,end) = 1;
net.view;
[net, tr] = train(net,tr_feature',tr_true');
% tr_feature is a 800x2 data matrix, tr_true is the 800x1 corresponding labels
Any idea?
Thanks in advance!
I've created a neural network to model a certain (simple) input-output relationship. When I look at the time-series responses plot using the nntrain gui the predictions seem quite adequate, however, when I try to do out of sample prediction the results are nowhere close to the function being modelled.
I've googled this problem extensively and messed around with my code to no avail, I'd really appreciate a little insight into what I've been doing wrong.
I've included a minimal working example below.
A = 1:1000; B = 10000*sin(A); C = A.^2 +B;
Set = [A' B' C'];
input = Set(:,1:end-1);
target = Set(:,end);
inputSeries = tonndata(input(1:700,:),false,false);
targetSeries = tonndata(target(1:700,:),false,false);
inputSeriesVal = tonndata(input(701:end,:),false,false);
targetSeriesVal = tonndata(target(701:end,:),false,false);
inputDelays = 1:2;
feedbackDelays = 1:2;
hiddenLayerSize = 5;
net = narxnet(inputDelays,feedbackDelays,hiddenLayerSize);
[inputs,inputStates,layerStates,targets] = preparets(net,inputSeries,{},targetSeries);
net.divideFcn = 'divideblock'; % Divide data in blocks
net.divideMode = 'time'; % Divide up every value
% Train the Network
[net,tr] = train(net,inputs,targets,inputStates,layerStates);
Y = net(inputs,inputStates,layerStates);
% Prediction Attempt
delay=length(inputDelays); N=300;
inputSeriesPred = [inputSeries(end-delay+1:end),inputSeriesVal];
targetSeriesPred = [targetSeries(end-delay+1:end), con2seq(nan(1,N))];
netc = closeloop(net);
[Xs,Xi,Ai,Ts] = preparets(netc,inputSeriesPred,{},targetSeriesPred);
yPred = netc(Xs,Xi,Ai);
perf = perform(net,yPred,targetSeriesVal);
figure;
plot([cell2mat(targetSeries),nan(1,N);
nan(1,length(targetSeries)),cell2mat(yPred);
nan(1,length(targetSeries)),cell2mat(targetSeriesVal)]')
legend('Original Targets','Network Predictions','Expected Outputs')
end
I realise narx net with a time delay is probably overkill for this type of problem but I intend on using this example as a base for a more complicated time-series problem in the future.
Kind regards, James
The most likely causes of poor generalization from the training data to new data is that either (1) there was not enough training data to characterize the problem, or (2) the neural network has more neurons and delays than are needed for the problem so it is overfitting the data (i.e. it is having an easy time memorizing the examples instead of having to figure out how they are related.
The fix for (1) is typically more data. The fix for (2) is to reduce the number of tap delays and/or neurons.
Hope this helps!
I'm not sure if you solved the problem yet. But there is at least one more solution to your problem.
Since you are dealing with a time series it is better (at least in this case) to set net.divideFcn = 'dividerand'. The 'divideblock' will only use the first part of the time series for training which may result in lost information about the long-term trends.
Increase the inputdelay, feedbackdelay and hiddenlayersize as following:
inputDelays = 1:30;
feedbackDelays = 1:3;
hiddenLayerSize = 30;
Also change function as
net.divideFcn = 'dividerand';
this changes work for me even though network take time