Comment by diyer22

Comment by diyer22 4 days ago

2 replies

In DDN, 1×1 convolutions are used only in the output layers of the Discrete Distribution Layer (DDL). The NN blocks between DDLs, which supply the fundamental computational power and parameter count, adopt standard 3×3 convolutions.

randomNumber7 4 days ago

Was there a specific reason for this choice?

  • diyer22 3 days ago

    1x1 convolution is the most lightweight operator for transforming features into outputs.

    3x3 convolution is the most common operator used to provide basic computational power.