fernandoandreotti / cinc-challenge2017 Goto Github PK

I know docker is very convenient, but my server was a docker. I found docker can't run in a docker, so I want to know if I don't use docker what requirements should I prepare or code I must change.

Thanks!

There is no main file

Hi,
How can i save the model after all cross validation sets are completed. How can i choose best validation set?

Hi Fernando
there is a problem in get_hrv function
HRV.NN50=sum(D>0.05); line 119
you have done the calculation in ms, so this should be abs(D)>50 I guess.
Iman

in readme you told the total number of features is 169. But when i run the extractfeatures.m file, i am getting 174 features.

Hi !
I have see some inverted datas in the ECG recordings.
Can you share the solution with me?

Is the paper available on the internet ? ("Comparing Feature Based Classifiers and Convolutional Neural Networks to Detect Arrhythmia from Short Segments of ECG")

Can you provide a testdata to estimate performance of the algorithm ? And can you also email me the preprint of this paper? Thanks a lot.

Hi I'm getting an error while running TrainClassifier.m during the step of crossvalidation loop 1,
the following is the error message.
"
NaN's cannot be converted to logicals.

Error in nntraining.fix_nan_inputs>iCopyNaN (line 134)
t{i}(isnan(yi)) = NaN;

Error in nntraining.fix_nan_inputs (line 57)
T = iCopyNaN(Y,T);

Error in nntraining.setup>setupPerWorker (line 104)
[X,Xi,Ai,T] = nntraining.fix_nan_inputs(net,X,Xi,Ai,T,Q,TS);

Error in nntraining.setup (line 43)
[net,data,tr,err] = setupPerWorker(net,trainFcn,X,Xi,Ai,T,EW,enableConfigure);

Error in network/train (line 335)
[net,data,tr,err] = nntraining.setup(net,net.trainFcn,X,Xi,Ai,T,EW,enableConfigure,isComposite);

Error in TrainClassifier (line 123)
net = train(net,In(trainidx,:)',Outbi(trainidx,:)');
"

When I try to load the model i get this error:

UnicodeDecodeError: 'utf-8' codec can't decode byte 0xe7 in position 30: invalid continuation byte

I already tried some solutions but all of them give the same error. Could you help me solve this?

When I use your model to predict the data, I make the following error:
Error in loading the saved optimizer '
'state. As a result, your model is '
'starting with a freshly initialized '
'optimizer.
How to solve it？
In addition, download a good training set, how to use your train.py for training

Hi,
In the second convolutional block you have in train_model.py line 208:
x1 = MaxPooling1D(pool_size=poolsize, strides=poolstr)(x1) and then the maxpooling for shortcut path.
But in the paper Rajpurkar et al. this maxpooling is not part of the architecture. (according to Fig 2. of their paper).
Would you please let me know how would it affect the training?!
Iman

Hi
Thanks for amazing toolbox you prepared.
HRbpm in in ExtractFeatures should be HR in bpm I assume?!
Since in the code :
HRbpm = median(60./(diff(qrsseg{end})));
The sampling frequency is missing!?

I ran ExtractFeatures() program. It created the following variable

allfeats_olap0.8_win10.mat

96229173 allfeats.mat
85282 - references.mat
allfeats.mat file should contain 8528173. But it contains 96229173. Dimension mismatch between the feature samples and the labels. Could you please explain why?

Hi，
I have implemented your DL method and used ResNet_30s_34lay_16conv.hdf5 model to classify all training data(8528), get mean F1 0.6117. But the result shows your model should get F1 0.92 on training data. I am very confused about the difference.

Attached the image for detailed information