Advances in Deep Learning for Time Series Forecasting and Classification: Winter 2023 Edition
The downfall of transformers for time series forecasting and the rise of time series embedding methods. Plus advances in anomaly detection, classification, and optimal (t) interventions.
Published in
16 min readJan 10, 2023
*Note you can find an updated 2024 issue of this article in DDS.
It has been quite sometime since I’ve written an update on the state of deep learning for time series. Several conferences have come and gone and the field as a whole has advanced in several different ways. Here I will attempt to cover some of the more promising as well as critical papers that have come out in the last year or so as well as updates to the Flow Forecast framework [FF].
Flow Forecast Framework updates:
- Over the last year we have made major strides in the architecture and documentation of FF. Just recently we rolled out full support for time series classification and supervised anomaly detection. Additionally, we have added several more tutorial notebooks and expanded our unit-test coverage to more than 77%.
- We also added a vanilla GRU model that you can use for time series…
