KDD-OpenSource / DeepADoTS

Repository of the paper "A Systematic Evaluation of Deep Anomaly Detection Methods for Time Series".
MIT License
569 stars 117 forks source link

The problem with allocating memory for training and predicting the DAGMM model #217

Open MartyanovVV opened 2 weeks ago

MartyanovVV commented 2 weeks ago

How to train a model if your data contains at least 500k sequences of 720 elements each. I tried to rewrite the fit and predict methods, but nothing worked out. Maybe there is some way to make fit run several times for one dataset, just breaking it into pieces and training on each of them? Also with the predict method.