Skip to content

This is my master thesis project. The goal is to apply dropout within RNN in order to obtain uncertainty for predictions.

Notifications You must be signed in to change notification settings

khanetor/dropout-in-rnn

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

36 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Dropout as Bayesian Approximation in RNN

Master thesis - Aalto University - Machine Learning and Data Science

Dataset

TODOs

  • Add biases
  • Make loss function
  • Stack multiple LSTM layers
  • Handle single dimension input
  • Iterate Stochastic pass once during training
  • Use parameter optimizer instead of writing custom loss functions
  • Make stochastic modules for regression
  • Make stochastic modules for classification
  • Compute predictive variance for classification
  • Compute predictive variance for regression
  • Optimize dropout rate (using Concrete Dropout)
  • Download clinical data
  • Early stop in training (use validation error)
  • Make equivalent in Tensorflow

About

This is my master thesis project. The goal is to apply dropout within RNN in order to obtain uncertainty for predictions.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published