A deep learning tool for time series classification and regression
Regression and classification on your time series data using automatic deep learning architecture selection.
mcfly 3.0 lets you train new network types on your time-series data. It brings support for ResNets and InceptionTime architectures next to the already supported CNN and DeepConvLSTM architectures. Version 3.0 also brings API improvements, bugfixes and uses Tensorflow 2.0 under the hood.
Add class weight support to work with unbalanced data sets.
This release makes the switch from working on top of Keras (with Tensorflow or Theano underneath) to working on top of Tensorflow directly. Because of this dependency change saved models with earlier versions are no longer compatible with the current version.
This release makes the switch from working on top of Keras (with Tensorflow or Theano underneath) to working on top of Tensorflow directly. Because of this dependency change saved models with earlier versions are no longer compatible with the current version.
Small patch that changes the requirements of the package to require keras version lower than the absolute lates 2.3.0 as this version is net yet supported by mcfly.
A release to fix citability and documentation format.
This release is a small patch over v1.0.2. It fixes configuration with Zenodo so the release gets assigned a DOI. Also broadened the support for versions of dependency: six.