Once https://github.com/leoniekerken/ECoG-foundation-model/pull/8 is submitted we will have an encoding downstream task implemented (word embedding -> neural embedding). We would also like to have a decoding downstream task that goes from neural embedding -> word embedding. This should be pretty simple to do by reusing the code from encoding and just reversing the direction of the regression.
Once https://github.com/leoniekerken/ECoG-foundation-model/pull/8 is submitted we will have an encoding downstream task implemented (word embedding -> neural embedding). We would also like to have a decoding downstream task that goes from neural embedding -> word embedding. This should be pretty simple to do by reusing the code from encoding and just reversing the direction of the regression.