I was able to achieve RMSE of 0.875 on the cross validation set with 3 features and lambda regularization parameter set to 2. After PCA (dimensionality reduction) to 2D (to visualize data), RMSE increased to 0.888.
Following image shows learning curves of this setting:
Following image visualizes movies in 2 dimensional projection:
I was able to achieve RMSE of 0.8662 on the cross validation set with 9 features and lambda regularization parameter set to 8. After PCA (dimensionality reduction) to 2D (to visualize data), RMSE increased to 0.909.
Following image shows learning curves of this setting:
Following image visualizes movies in 2 dimensional projection:
