search

liyaguang / DCRNN

Implementation of Diffusion Convolutional Recurrent Neural Network in Tensorflow
MIT License
1.22k stars 400 forks source link

What is the meaning of the last dimension of the input (Batch_size, Time granularity, Node number,2) of the network? #19

Closed zhengxuyu closed 5 years ago

zhengxuyu commented 5 years ago

Hi, where can I get the introduction of the METR-LA and the PEMS-Bay dataset? What is the meaning of the last dimension of the input (Batch_size, Time granularity, Node number,2) of the network?

liyaguang commented 5 years ago

Hi zhengxu, the input data is (N x T x V x D), where N is the number of training examples, T corresponds to the number of input steps, V denotes the number of nodes, and D represents the input feature dimension, here 2 consists of 1) the speed and 2) the time.

zhengxuyu commented 5 years ago

About the second dim, what 'the time' means to be specific?

yansicing commented 5 years ago

The second dimension T is timesteps, which means seq_len, whose value is 12 The last dimension D is the speed corresponding to a certain time ~

mengmeng716 commented 5 years ago

Hi, liyaguang @liyaguang thanks for sharing the great work. actually I don't think the second input feature dimensions is time. because I find the value is something like float in In scientific counting, the value is between [0,1] , mean value 0.497, max value 0.997, min value 0; so can you provide some evidence for the second feature dimenson please?

thanks and best regards,

chenz97 commented 4 years ago

Hi @yansicing , thanks for your explaination, but I still have questions here. You said "D is the speed corresponding to a certain time", and I'm wondering which time it is. I was thinking the authors were using 5-minutes average speed, right? If so, what is the time that the speed corresponds to? Thank you very much!

liyaguang commented 4 years ago

Hi @chenz97, @mengmeng716, the time dimension refers to the normalized time in a day, e.g., 0:00 am will be 0, while 12pm will be 0.5.

chenz97 commented 4 years ago

Hi @liyaguang , I got it. Thanks a lot.