Dl4jMlpClassifier now supports resuming of training after user-specified number of epochs has completed (i.e. the network can be told to iterate for another x epochs). This allows serialized models to continue training at a later date. GUI/command-line support for this in Weka will be in 3.8.4/3.9.4 when released (and is available in nightly snapshots until that time). Also fixed command-line option handling for LogConfiguration.
Thanks! Are there any prebuilt nightly snapshots of weka available or could you point me to some build instructions? I'd like to add some test cases first.
Dl4jMlpClassifier now supports resuming of training after user-specified number of epochs has completed (i.e. the network can be told to iterate for another x epochs). This allows serialized models to continue training at a later date. GUI/command-line support for this in Weka will be in 3.8.4/3.9.4 when released (and is available in nightly snapshots until that time). Also fixed command-line option handling for LogConfiguration.