Solution for EEG Classification via Multiscale Convolutional Net coded for NeuroHack at Yandex.
Refactoring is in progress, this code is not entirely readable.
- Multifilter Convnet: Apply 1D Convolutions of various filter size to input timeseries and concatenate outputs to obtain the embedding
- Multiscale Convnet: Apply 1D Convolution and Global Max Pooling to input timeseries resampled at different scales and concatenate outputs to obtain the embedding