Feature/rf n classes - Githubissues

beckstev / MachineLearningSeminar

MachineLearningSeminar SS19 TU Dortmund

MIT License

0 stars 0 forks source link

Feature/rf n classes #44

Closed beckstev closed 5 years ago

beckstev commented 5 years ago

Features:

Added EarlyStopping to autoencoder
Directory of autoencoder now includes the number of classes
Added a GritSearch for the random forest
Added function to create the height width disttribution of our dataset
some minor changes

HowToTest: python train_autoencoder.py <encoder model> <image_resize> <n_classes> --gird_search

Be aware that the current gridsearch takes about 1 hour to finish!