DMVST-Net
DMVST-Net copied to clipboard
huaxiuyao
Dataset Format
Hi, I would like to inquire about the dataset format so when I have a similar dataset, I can actually transform it to match your specification. It would be helpful if you explain more about the data format. Thank you
Same question here. My inputs are as follow:
- seq_len = 8, feature_len = 16, local_image_size = 7, toponet_len = 32
- trainX = {ndarray: (*, seq_len, feature_len)}
- trainY = {ndarray: (*, 1)}
- trainimage = {ndarray: (*, seq_len, local_image_size, local_image_size, 1)} since there is only one channel in my local map
- traintopo = {ndarray: (*, toponet_len)}
Where * stands for total number of data generated from original dataset. Every value in each ndarray is ndfloat in range(0,1). Model could start fitting however the output values are incorrect. Is there a problem with my inputs?
Problem solved, I forgot to change those zeros in hyperparameters. @gunarto90 if you are still interested in this work, you could reference my data format. If you read the paper, here is a dictionary for you to understand what these variables stand for:
-
seq_len: $h$ -
feature_len: $|\mathbf{s}^i|$ -
toponet_len: $|\mathbf{m}^i|$ -
trainX: $\mathbf{s}^i$ -
trainY: $y^{i}_{t+1}$ -
trainimage: $\mathbf{Y}^i$ -
traintopo: $\mathbf{m}^i$
