Using 1d CNN for time series classification of patchy time series I would like to apply 1d CNN like models for the patchy time series classification. For a LSTM like model, my plan to deal with the patchiness problem is to add a time column for the data. However, it not clear to me how to add this into the 1d CNN as the convolution should attempt to pre-process signals for the data, but I wouldn't expect it to do this for time.
Has anyone tried this before, any idea of how to input time into a 1d CNN?
 A: Yes, people have tried this before. Please have a look at the following papers (I suggest in that order):

*

*Time Series Classification from Scratch with Deep Neural Networks: A Strong Baseline (https://arxiv.org/abs/1611.06455) [source code: https://github.com/cauchyturing/UCR_Time_Series_Classification_Deep_Learning_Baseline]

*Deep learning for time series classification: a review (https://arxiv.org/abs/1809.04356) [source code: https://github.com/hfawaz/dl-4-tsc]

More advanced methods (like transfer learning) on top of 1-D CNNs are discussed in the following papers:
3.) Timage -- A Robust Time Series Classification Pipeline (https://arxiv.org/abs/1909.09149) [source code: https://github.com/patientzero/timage-icann2019]
4.) InceptionTime: Finding AlexNet for time series classification (https://link.springer.com/article/10.1007/s10618-020-00710-y) [source code: https://github.com/hfawaz/InceptionTime]
Please also have a look at the corresponding repositories containing the source code for each approach.
