pandas-ml-utils-torch 0.2.7

Augment pandas DataFrame with methods for machine learning

pytorch extension for pandas-ml-utils

Adds a PytorchModel to the pandas ml utils suite. While a regular class extending nn.Module is sufficient, there is also a special class PytorchNN which can be extended as well. Using this class has the following advantages:

• allows to use L1, L2 regularization -> example
• allows different forward path for training and prediction (useful i.e. for reparameterization trick) -> example
• allows to implement auto-encoders easily by just providing the encode/decode functions
• added loss functions like SoftDTW (fit time series) loss or HeteroscedasticityLoss (fit Normal Distribution) -> example