🧵4/N
That experiment led me back to the HMS dataset and raised another question:
Could explicit temporal alignment and duration modeling be adapted to EEG spectrogram classification?
It would not be a direct transfer. EEG signals do not contain the same discrete
🧵1/N
One experience that asserted ML was the right field for me came in 2024, when I became obsessed with Kaggle’s HMS: Harmful Brain Activity Classification.
The challenge involved classifying raw EEG recordings and their corresponding spectrograms into 6 categories: seizure,
