jiang-yuan / learning-routing-dro Goto Github PK

Hello, author! I enjoy reading your article, but encounter a problem when running your code and found no data. Could you please provide a data file? Thank you very much if you can!

Thanks for your paper "Learning to Solve Routing Problems via Distributionally Robust Optimization", which is an interesting direction.

However, could you please double-check the provided source code?

For example, there doesn't seem to be any data.data module in this repository, or is it a dependency?
https://github.com/jiang-yuan/Learning-routing-DRO/blob/main/run.py#L9

Also, it would be much more helpful if you can provide a few demos for reproducing the training/evaluating processes in the paper (currently there is only a python3 run.py).

Hi, thank you for the great work.

I wanted to ask if you could provide the model checkpoints used in the evaluation of the paper together with the code,
such that one can run them as a baseline?

Your help is much appreciated.

How one-dimensional convolution works on a node and its k-nearest neighbors?

Specifically, given a node $i$ and its k-nearest neighbors $i_1,\ldots i_k$. As stated in the paper, the embedding $h_i$ is calculated by $h_i=W_1 x_i+W_2 \bar{h_i}$, where $x_i$ is the coordinate of node $i$, $\bar{h_i}$ is the convolutional results of CNN embedding layer s, and $W_1$ and $W_2$ are trainable matrices. I wonder how is $\bar{h_i}$ calculated. Thanks!

Thanks for your paper on generalization about routing problems, which helps me a lot.

But I found that the code for data.data is missing. Could you please provide it? This will help me better understand the paper. Thank you very much