A Clockwork RNN model to train on compressional velocity (Vp), formation density (Rhob), gamma ray (Gr), and resistivity (Rt) logs to predict Neutron Porosity (Nphi)
The clockwork RNN model is a tensorflow implementation of Koutnik et al., 2014.
The modifications include:
- Fixed mask of the hidden layer
- Added correct selection of block-rows of the hidden layer for evaluation
- Added prediction point-by-point
- Added training and validation losses plots
- Added learning rate decay plot