we provided dataset mentioned in our paper in google drive.
you can see it using python with numpy package.
train_data = np.load("X_train.npy").squeeze(2)
print(train_data)
print(train_data.shape)
[[ 0.92631579 0.92982456 0.92631579 ... 0.78947368 0.03508772
0.47017544]
[ 0.92982456 0.92631579 0.92982456 ... 0.03508772 0.47017544
0.92982456]
[ 0.92631579 0.92982456 0.92982456 ... 0.47017544 0.92982456
0.98596491]
...
[-0.86315789 -0.85614035 -0.87719298 ... -0.87719298 0.32631579
-0.57192982]
[-0.85614035 -0.87719298 0.32631579 ... 0.32631579 -0.57192982
-0.87719298]
[-0.87719298 0.32631579 -0.03157895 ... -0.57192982 -0.87719298
0.32631579]]
(158373, 100)
as we can see , the first output the specific content ,the second output show the format ,the dataset is composed of 158373 piece of 100 dim data,each data range in (-1,1).
we also provided the chord dict we used in our paper in dictionary './dataset' that the first column is the index of the chord and the second column is the chord itself.