Using the latent space of a variational autoencoder to perform symbolic regression by sampling equations.
Project description
Variational Autoencoder Embeddings Symbolic Regression (VAEESR)
This python package performs Symbolic Regression by creating an Embedding where semantically similar equations are close to each other and it uses MCMC sampling to find the equation that is the closest to the oberved data. Therefore, is uses three main functions:
create_dataset which creates a customizable dataset that can be ajusted for the specific problem at hand. The main parameters are: The x_values for which the functions are evaluated, the range of constants, the maximum tree depth, the possible operators and functions and the total number of equations in the dataset.
create_autoencoder which trains an autoencoder with the dataset. The some hyperparameters can be ajdusted as well.
perform_MCMC which performs the symbolic regression by sampling equations from the autoencoder embedding.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.