DL backlog - model comparison

[martins0n]

DL models backlog to be implemented:

name	TLDR	priority
Autoformer	current best $\mathcal{O}( N \log N)$ transformer. Long horizon forecasts
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		🔥

Metrics
(There are few intersections or authors reused the same metrics from other papers without reproduced results)