View on Github

Welcome to theta’s documentation!

Theta is a machine learning (ML) framework implementing the Riemann-Theta Boltzmann Machine (RTBM), written in Python and Cython. It offers a high-level interface to build and train RTBM based ML architectures for probability density estimation, data regression and classification.

The code implements the RTBM as described in the theoretical paper arXiv:1712.07581.

Note

Theta is in a proof-of-concept / research phase. You may observe that model training requires some fine tune to get proper results.

User documentation

• Getting started
  • Installation
  • Testing Installation
• Components
  • RTBM
  • Model
  • Layers
    • Theta Probability Unit
    • Theta Diagonal Expectation Unit
    • Normalized Additive
    • Linear
    • Non-Linear
  • Minimizers
    • Evolutionary algorithm
    • Gradient descent
    • Gradient descent schemes
  • Activations
    • Linear
    • Sigmoid
    • Tanh
  • Initializers
    • Uniform
    • Normal
    • Null
    • Glorot normal
    • Glorot uniform
  • Cost functions
    • MSE
    • Logarithmic
    • Sum
    • RMSE
  • Stopping conditions
    • Early Stop
• Examples and tutorials
  • Probability estimation
  • Data regression and classification

License and citation policy

The theta package is an open-source package under AGPLv3. If you use the theta package please cite the following article:

Daniel Krefl, Stefano Carrazza, Babak Haghighat, Jens Kahlen.
Riemann-Theta Boltzmann Machine, arXiv:1712.07581