search

lanxiang1017 / DynamicBadPairMining_ICLR24

DBPM is a simple algorithm designed as a lightweight plug-in without learnable parameters to enhance the performance of time series contrastive learning.
https://openreview.net/pdf?id=K2c04ulKXn
12 stars 2 forks source link

Clarification on Data Processing #1

Closed ShijianXu closed 6 months ago

ShijianXu commented 6 months ago

Dear authors,

Thank you for sharing the code for your work. I am new to ECG processing, so please excuse any novice questions. In reviewing the PTB-benchmark paper, I noticed they utilized a sliding window approach to extract random ECG segments, such as 2.5s, as input. However, upon reviewing their code, I couldn't identify this specific processing step. Your code is based on their repo, yet I couldn't find this processing either. Could you please clarify how your data is shaped before inputting it into the model?

Thank you very much for your assistance! Best regards.

lanxiang1017 commented 6 months ago

Hi Shijian,

We only use the original 100Hz ECG data, not their classification models with a sliding window strategy. Therefore, the input data shape is 12 * 1000, where 12 represents the number of leads and 1000 represents the length of the sequence.

I hope this clarifies.

ShijianXu commented 6 months ago

Thanks for the clarification!