kristpapadopoulos / seriesnet Goto Github PK

I don't understand something you are making output as (NONE,length,1) but you take only the last point and your y_output data is only one point shifted from your input data for example:

x=[0,1,2,3,4,5,...63]
y=[64]

why not making only a flatten layer and dense layer for getting the output of this point ?
will it make a big difference ?

That's a great architecture you have there. I wanted to know if your paper has been published? so, that I may add to references of my paper. Thanks.

This sounds pretty interesting, but I get this when I try to run it on a simple 1D array:
Traceback (most recent call last): File ".\seriesnet.py", line 130, in <module> print(evaluate_timeseries(np.arange(100),10)) File ".\seriesnet.py", line 107, in evaluate_timeseries model = DC_CNN_Model(length) File ".\seriesnet.py", line 67, in DC_CNN_Model l1a, l1b = DC_CNN_Block(32,2,1,0.001)(input) File ".\seriesnet.py", line 56, in f network_out = merge([residual, network_in], mode='sum') TypeError: 'module' object is not callable

Can you please provide an example of how to run the code?
I tried:
a = np.array([0,1,2,3,4,5]) evaluate_timeseries(a , 1)
but it doesn't work.

Hey, thanks for this great code! As for saubersf running a simple example does not work for me. I get an error for the model.fit(X,y) call in line 127. Error message is "ValueError: None values not supported." Do you know how to solve this error?

HI,

I have a question concerning the input of several time series to exploit correlations between them, can I just change the shape in:

def DC_CNN_Model(length):

input = Input(shape=(length,#OF TS IN THE INPUT))

Thank you very much.

hi，
I don't understand why the batch of this network is set to 1.

Hi,
I have tried to test your implementation of wavenet for energy production data.
The input looks like :

time_serie = [257,247,244,245,231, ...]
test_size = 24

I did :
predictions = evaluate_timeseries(time_serie, test_size)
but predictions are values between -0.1 and 0.

Do you know what could be the reason ?

Thanks a lot

Dear Krist, can I invite you to contribute into the following repository? https://github.com/ZhengyaoJiang/PGPortfolio