Few more questions on SAITS working method
Rajesh90123 opened this issue · comments
I am really impressed with your work: "SAITS: SELF-ATTENTION-BASED IMPUTATION FOR TIME SERIES".
I was studying your code and paper and I have few questions:
- Is there any particular reason for not using learning rate scheduler?
- Also, when I tried replicating the code, I found out that my program was running indefinitely during random search for hyperparameter tuning especially with the use of loguniform in learning rate. I didn't find any lines regarding maximum number of trials for hyperparameter tuning in your code or in your paper "SAITS: SELF-ATTENTION-BASED IMPUTATION FOR TIME SERIES". Could you please provide information regarding this?
- Do you change the value of artificial missing rate for MIT in training section based on missing rate at test? What I mean here is that, for example, if you have to test your model on test dataset with 80% missing rate, will the MIT missing rate be fixed to 20% as you have done in the code provided or do you train the entire model again by changing MIT missing rate manually to 80% in the training code?
- Do you perform hyperparameter tuning on different missing rate in validation data or do you perform hyperparameter tuning on a particular missing rate at validation dataset and save it to use it for test dataset with any missing rate? What I mean here is, running hyperparameter tuning on a validation dataset of 20% missing rate and saving the best model, and using the same best model to impute missing data for test dataset with 80% missing rate?
Thank you, and if my questions are not clear, please comment, and I will try my best to describe my question. English is my second language, and I am not fluent on speaking English.
Hi Rajesh, thanks for your attention to our work!
- I'd like to ask, do you have any particular reason to use a scheduler? This is not an LLM, SAITS is not hard to train. But if you prefer to give it a try, you can use PyPOTS and optimizers in it support lr scheduling;
- No, I didn't set a maximum num of trials. I just let the experiment run til I thought it was good to stop;
3&4. The answers are both No. I've written everything I did and found in the paper. If you didn't find the answer, you should perform experiments to do it yourself. Try it if you think it's worth a shot. BTW, here is our new paper about the masking strategies https://arxiv.org/abs/2405.17508, and it may be helpful to you.
Please kindly cite our papers in your work if you think they're useful ;-)
Thank you so much sir for your response 🙏.
No problem. If you work with time series, follow PyPOTS https://pypots.com closely. 🤗