Alignment of base model output and get_model_input about eeg_classification

[CVxTz]

Hello,
The input of get_base_model is (3000, 1) and its output is (64,) , this model should not be used on its own but as part of get_model_cnn.
The input of get_model_cnn is (None, 3000, 1), the submodel that outputs (64, ) is only part of it.

[CVxTz]

Go here for an explanation : https://towardsdatascience.com/sleep-stage-classification-from-single-channel-eeg-using-convolutional-neural-networks-5c710d92d38e

[zizi96]

Ok so I think the issue is with my training input shape, I'm still fairly new to ML.

Am i correct in assuming this code taken from the run-file feeds training samples (of shape 3000,1) into base_model 1 at a time?

train_val, test = [x for x in files if x.split("/")[-1][:5] in train_ids],\
                  [x for x in files if x.split("/")[-1][:5] in test_ids]

train, val = train_test_split(train_val, test_size=0.1, random_state=1337)

train_dict = {k: np.load(k) for k in train}
test_dict = {k: np.load(k) for k in test}
val_dict = {k: np.load(k) for k in val}

[CVxTz]

No its wrong.
You need to checkout the function "gen" in https://github.com/CVxTz/EEG_classification/blob/master/code/utils.py

[zizi96]

I have an idea but its only a guess. I tried generating the data myself to check myself but I don't use Linux. Its very difficult to visualise what gen is doing without knowing exactly what the data into it looks like hence why I would greatly appreciate knowing the shape of what gen outputs.
From there I could learn whats going on with pre-processing via reverse engineering. Thanks for your time.

[CVxTz]

gen outputs X, Y.
X shape is (batch_size, 100, 3000, 1)
Y shape is (batch_size, 100, 1)
You can double check by calling :
X, Y = next(gen(train))
print(X.shape)
print(Y.shape)