WenjieDu/PyPOTS

"""
The core wrapper assembles the submodules of TiDE imputation model
and takes over the forward progress of the algorithm.
"""

# Created by Wenjie Du <wenjay.du@gmail.com>
# License: BSD-3-Clause

import torch.nn as nn

from ...nn.modules.saits import SaitsLoss, SaitsEmbedding
from ...nn.modules.tide.autoencoder import TiDE


class _TiDE(nn.Module):
    def __init__(
        self,
        n_steps,
        n_features,
        n_layers,
        d_model,
        d_hidden,
        d_feature_encode,
        d_temporal_decoder_hidden,
        dropout,
        ORT_weight: float = 1,
        MIT_weight: float = 1,
    ):
        super().__init__()

        self.n_steps = n_steps
        self.saits_embedding = SaitsEmbedding(
            n_features * 2,
            d_model,
            with_pos=False,
            dropout=dropout,
        )
        # n_output_steps = n_steps
        # n_output_features = n_features
        # d_flatten = n_steps + (n_steps + n_output_steps) * d_feature_encode
        # self.feature_encoder = ResBlock(
        #     n_features,
        #     d_hidden,
        #     d_feature_encode,
        #     dropout,
        # )
        # self.encoder = TideEncoder(
        #     n_steps,
        #     n_features,
        #     n_layers,
        #     d_flatten,
        #     d_hidden,
        #     dropout,
        # )
        # self.decoder = TideDecoder(
        #     n_steps,
        #     n_steps,
        #     n_output_features,
        #     n_layers,
        #     d_hidden,
        #     d_feature_encode,
        #     dropout,
        # )
        # self.temporal_decoder = ResBlock(
        #     n_output_features + d_feature_encode,
        #     d_temporal_decoder_hidden,
        #     1,
        #     dropout,
        # )
        # self.residual_proj = nn.Linear(n_steps, n_output_steps)
        self.tide = TiDE(
            n_steps,
            n_features,
            n_layers,
            d_hidden,
            d_feature_encode,
            d_temporal_decoder_hidden,
            dropout,
        )

        # for the imputation task, the output dim is the same as input dim
        # self.output_projection = nn.Linear(d_model, n_features)
        self.saits_loss_func = SaitsLoss(ORT_weight, MIT_weight)

    def forward(self, inputs: dict, training: bool = True) -> dict:
        X, missing_mask = inputs["X"], inputs["missing_mask"]

        # # WDU: the original TiDE paper isn't proposed for imputation task. Hence the model doesn't take
        # # the missing mask into account, which means, in the process, the model doesn't know which part of
        # # the input data is missing, and this may hurt the model's imputation performance. Therefore, I apply the
        # # SAITS embedding method to project the concatenation of features and masks into a hidden space, as well as
        # # the output layers to project back from the hidden space to the original space.
        # X_enc = self.saits_embedding(X, missing_mask)
        #
        # # TiDE encoder processing
        # feature = self.feature_encoder(missing_mask)
        # # print(f"X.shape: {X.shape}")
        # # print(f"feature.shape: {feature.shape}")
        # enc_in = torch.cat([X_enc, feature], dim=-1)
        # enc_out = self.encoder(enc_in)
        # dec_out = self.decoder(enc_out)
        # reconstruction = self.temporal_decoder(
        #     torch.cat([feature[:, self.n_steps :], dec_out], dim=-1)
        # ).squeeze(-1) + self.residual_proj(X_enc)
        # # # project back the original data space
        # # reconstruction = self.output_projection(enc_out)
        reconstruction = self.tide(X, missing_mask)

        imputed_data = missing_mask * X + (1 - missing_mask) * reconstruction
        results = {
            "imputed_data": imputed_data,
        }

        # if in training mode, return results with losses
        if training:
            X_ori, indicating_mask = inputs["X_ori"], inputs["indicating_mask"]
            loss, ORT_loss, MIT_loss = self.saits_loss_func(reconstruction, X_ori, missing_mask, indicating_mask)
            results["ORT_loss"] = ORT_loss
            results["MIT_loss"] = MIT_loss
            # `loss` is always the item for backward propagating to update the model
            results["loss"] = loss

        return results