Is the target month of nino3.4 right?

[Jane-2808140]

When I was running the code, I found that there is a problem in the function "reformat_cnn_data" in the file "data_wrangling.py". The comment in the code says that the target month is [25, 36), which is consistent with what you wrote in your paper and the CNN paper, but in the code, when lead month is 1, windows is 3, and target month is 0, the preditor months is [22,25), If it is written according to the paper, the target month should be 25, that is, 1 instead of 0 (24) in the code. I wonder if my understanding is correct, or I did not understand the calculation method of nino3.4 index moving average in the data set. I hope you can reply me as soon as possible after reading it. Thank you！

[salvaRC]

Hi! The following is true: "when lead month is 1, windows is 3, and target month is 0, the preditor months is [22,25)" . But I'm not sure what you mean by "target month should be 25, that is, 1 instead of 0 (24) in the code" ... Can you point to the code lines that you find hard to understand? Also, please note that this code lines are a slightly refactored version of the data loading code from the paper we took the data from. E.g. check https://github.com/jeonghwan723/DL_ENSO/blob/master/CNN/sample/nino34.train_cmip.sample