Solution for the Quora question pairs competition on kaggle
This is my solution to this competition https://www.kaggle.com/c/quora-question-pairs on kaggle. I have tried couple of models which includes the following. I will be adding all the approaches sequentially.
Manhattan LSTM Model (Train Accuracy - 85% ,Train loss - 0.1, Validation Accuracy - 80%, Validation Loss - 0.14)
Siamese 2 hidden LSTM Layers with 3 Dense Layers(Train-Accuracy - 86%,Train-loss- 0.31 Validation-Accuracy - 82%,Validation-Loss- 0.4)
Siamese Dense model with 3 Dense Layers (Train-Accuracy - 88%,Train-loss- 0.29)