search

WenjieDu / PyPOTS

A Python toolkit/library for reality-centric machine/deep learning and data mining on partially-observed time series, including SOTA neural network models for scientific analysis tasks of imputation/classification/clustering/forecasting/anomaly detection/cleaning on incomplete industrial (irregularly-sampled) multivariate TS with NaN missing values
https://pypots.com
BSD 3-Clause "New" or "Revised" License
1.01k stars 95 forks source link

allow customized loss function for imputation/informer #432

Open mmo139 opened 3 months ago

mmo139 commented 3 months ago

1. Feature description

Add the customized_loss_func feature currently available in SAITS to other imputation models.

2. Motivation

Currently saits_loss_func is used in class _Informer (line 96 of pypots/imputation/informer/core.py), indicating that loss function is MAE and cannot be customized with a model parameter (the original paper used MSE). I adjusted locally for my task for using MSE but it would be nice if customized_loss_func can also be added for informer. Besides, it seems some other models are using saits_loss_func in core.py as well (but I didn't check one by one).

3. Your contribution

I can help with updating core.py / model.py but not sure if rolling out the feature for other models are also desired.

github-actions[bot] commented 3 months ago

Hi there 👋,

Thank you so much for your attention to PyPOTS! You can follow me on GitHub to receive the latest news of PyPOTS. If you find PyPOTS helpful to your work, please star⭐️ this repository. Your star is your recognition, which can help more people notice PyPOTS and grow PyPOTS community. It matters and is definitely a kind of contribution to the community.

I have received your message and will respond ASAP. Thank you for your patience! 😃

Best, Wenjie