issue about Input of TCN

locuslab / TCN

Sequence modeling benchmarks and temporal convolutional networks

https://github.com/locuslab/TCN

MIT License

4.17k stars 879 forks source link

issue about Input of TCN #62

Closed moelmadoka closed 3 years ago

moelmadoka commented 3 years ago

Hi,I would like to understand the meaning of the input (B, C, L) of the TCN model. B represents the batch size. Does C represent the dimension of each sample feature? Does L represent the sequence length of each sample? Does args.nhid in the network represent the convolution of the length of the sample sequence? For multivariate time series modeling, what input should tcn handle? For example, data.shape is (data_number, feature_dim).

jerrybai1995 commented 3 years ago

Your interpretation of C and L is correct.

For multivariate time series modeling, you should use it similar to how you use RNN/LSTM. Make sure your data has a temporal/sequence-length dimension.