Companion repository to "On a time-frequency blurring operator with applications in data augmentation"

This repository contains code to repeat the main experiment of the paper. Call download_and_save_data() to download the SpeechCommands V2 dataset. All required Python packages should be self-explanatory. The improve() function works on the accs pickle file to add more examples of accuracies. Once training is complete, the export_results() prints accuracies and standard errors in a LaTeX-friendly way.