i have question in the Nhits model in the paper in multi rate data sampling does the model consider the low frequency and high frequency in different block or just the low frequency in sampling step before the Maxpool?
07/20/2023, 1:51 PM
Hi @marah othman! Each block is specialized in a different frequency. To do that, the sampling rate (controlled with the kernel size of the maxpool payer), is different across blocks. High frequency blocks with have a shorter kernel (less pooling) to preserve more information. Hope this helps!