shift-invariant

pip install antialiased-cnns to improve stability and accuracy

Repository has been moved: https://github.com/adobe/antialiased-cnns

The official implementation of ICLR2025 paper for shift-invariant neural networks

In this research project we used a shift-invariant k-means algorithm to learn a preictal and interictal codebook of prototypical waveforms that can be used to summarize the occurrence of recurrent waveforms and to classify between preictal and interictal segments. We use the common spatial patterns (CSP) method to spatially filter the multichannel ECoG segments before doing the waveform learning.