You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
ConvShift Layer is designed to implement circular convolutional operation. Circular convolutional operation is a necessary step to implement location-based addressing in Neural Turing Machine. NTM was designed to process sequence data initially. However, current ConvShift Layer only supports non-sequence type data input.
Original implementation calls circularConv and circularConvDervative functions implemented in class Matrix.
First, I will check the type of input data. If the input data is non-sequence type, the function will run origin logic, otherwise call circularConvSeq and circularConvSeqDerivative.
ConvShift Layer is designed to implement circular convolutional operation. Circular convolutional operation is a necessary step to implement location-based addressing in Neural Turing Machine. NTM was designed to process sequence data initially. However, current ConvShift Layer only supports non-sequence type data input.
Original implementation calls circularConv and circularConvDervative functions implemented in class Matrix.
First, I will check the type of input data. If the input data is non-sequence type, the function will run origin logic, otherwise call circularConvSeq and circularConvSeqDerivative.
Please go to related PR #2133 to check implementation details.
The text was updated successfully, but these errors were encountered: