Below are few of the models I tried and the corresponding MSE Losses for those models.
| Architecture | MSE loss |
|---|---|
| NN with only Linear and Tanh layers | ~0.021 |
| NN with Linear, Tanh and BatchNorm layers | 0.0161 |
| NN with Linear, Tanh and BatchNorm layers, and with principal components concatenated to the Data | 0.00774 |
| NN with Linear, Tanh and BatchNorm layers, and with singular values concatenated to the Data | 0.00706 |