Skip to content
/ Neuvol Public

Genetic algorithm framework to find the best neural network architecture with PyTorch

License

Notifications You must be signed in to change notification settings

Qwinpin/Neuvol

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

60 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

logoNeuvol

Genetic algorithm to find the best neural network architecture with Keras

Neuvol is a genetic algorithm API for neural networks based on Keras. The main idea is to work with data only, without direct architecture construction (but it is still possible).

Grow your own neural network!
  • Data in -> Neural Network Architecture out
  • Set your assumptions about architecture, it will be taken into account! (not yet)
  • A large number of allowed layers types

Features:

  • Supported data types: texts
  • CNN, Dense, LSTM layers are available
  • Dropout and reshape sub-layers (Flattern) are available too

Getting started:

TODO

TODO

  • Architectures distribution generation
  • Images support
  • Regression models
  • Generative networks (??? almost impossible)
  • More available layers
  • More options (assumptions)
  • Logo
  • Visualization