Viet-Anh on Software Logo

What is: Causal Convolution?

SourceWaveNet: A Generative Model for Raw Audio
Year2000
Data SourceCC BY-SA - https://paperswithcode.com

Causal convolutions are a type of convolution used for temporal data which ensures the model cannot violate the ordering in which we model the data: the prediction p(xt+1x1,,xt)p(x_{t+1} | x_{1}, \ldots, x_{t}) emitted by the model at timestep tt cannot depend on any of the future timesteps xt+1,xt+2,,xTx_{t+1}, x_{t+2}, \ldots, x_{T}. For images, the equivalent of a causal convolution is a masked convolution which can be implemented by constructing a mask tensor and doing an element-wise multiplication of this mask with the convolution kernel before applying it. For 1-D data such as audio one can more easily implement this by shifting the output of a normal convolution by a few timesteps.