search

emadeldeen24 / TS-TCC

[IJCAI-21] "Time-Series Representation Learning via Temporal and Contextual Contrasting"
MIT License
357 stars 99 forks source link

About Unsupervised Downstream Tasks #34

Closed kke12580 closed 3 months ago

kke12580 commented 4 months ago

Dear author, thank you for your contribution. I’ve been using your code and noticed that it seems unsuitable for unsupervised downstream tasks. Specifically, my downstream task involves using an autoencoder for binary class anomaly detection. If it’s feasible, could you inform me which part of the code should be modified?

emadeldeen24 commented 3 months ago

you can use the autoencoder itself to perform the pretraining by doing an unsupervised reconstruction task (you can follow MAE works). Then use it for the downstream task.