Description

1. Why the change?

Refer to #6405
Inspired by the methods mentioned in Are Transformers Effective for Time Series Forecasting?, adding new extensions to tcn model. And the experiment results on the three datasets (electricity, nyc_taxi and others) demonstrate the effectiveness of the added extensions.

2. User API changes

• Add new options "dummy_encoder", "normalization" and "decomposition_kernal_size" to bigdl.chronos.forecaster.tf.tcn_forecaster.TCNForecaster.init()
• Add new options "normalization" and "decomposition_kernal_size" to bigdl.chronos.forecaster.tf.seq2seq_forecaster.Seq2SeqForecaster.init()
• Add new options "normalization" and "decomposition_kernal_size" to bigdl.chronos.forecaster.tf.lstm_forecaster.LSTMForecaster.init()

:param dummy_encoder: bool, no encoder is applied if True, which will turn TCNForecaster to a Linear Model. If True, input_feature_num should equals to output_feature_num.

:param normalization: bool, Specify if to use normalization trick to alleviate distribution shift. It first subtractes the last value
of the sequence and add back after the model forwarding.

:param decomposition_kernel_size: int, Specify the kernel size in moving average. The decomposition method will be applied if and only if decomposition_kernel_size is greater than 1, which first decomposes the raw sequence into a trend component by a moving average kernel and a remainder(seasonal) component. Then, two models are applied to each component and sum up the two outputs to get the final prediction. This value defaults to 0.

3. Summary of the change

Add dd new tricks for tf forecasters.

4. How to test?

• N/A
• Unit test
• ...