Etna-native neural network models
No due date
85% complete
DL models backlog to be implemented:
| name | TLDR | priority |
|---|---|---|
| Autoformer | current best |
|
| N-Beats | Uncommon net architecture. Beats es-rnn(M4 winner) on M4 | 🔥 |
| N-Hits | N-Beats predecessor. Beats Autoformer. Long horizon forecasts | 🔥 🔥 |
| TFT | Interpretable time series forecasting. Transfromer bases. … |
DL models backlog to be implemented:
| name | TLDR | priority |
|---|---|---|
| Autoformer | current best |
|
| N-Beats | Uncommon net architecture. Beats es-rnn(M4 winner) on M4 | 🔥 |
| N-Hits | N-Beats predecessor. Beats Autoformer. Long horizon forecasts | 🔥 🔥 |
| TFT | Interpretable time series forecasting. Transfromer bases. Probabilistic forecasts | 🥶 |
| DeepAR | RNN based. Probabilistic forecasts | 🥶 |
| NLinear | Beats Autoformer. Simple | 🔥 |
| Vanilla Transformer | 🔥 | |
| DeepTime | 🔥 | |
| TIDE | 🔥 | |
| TimesNet | 🔥 🔥 🔥 | |
| Metrics | ||
| (There are few intersections or authors reused the same metrics from other papers without reproduced results) |