In the context of the residual learning framework, the primary purpose of introducing skip connections is to enable the network to learn the __________ of the desired output with respect to its inputs.

In residual learning, skip connections let the network focus on learning the difference between the target output and a simple approximation of it, rather than the whole output. By adding the input directly to the output of a few layers, the network is asked to model only the missing part, or residual, that the layers need to learn. This simplifies training because the residual is often smaller and easier to represent. For example, if an input image has a bright region, the network can learn only the extra brightness that a simple identity mapping does not capture, leaving the rest to the skip path. Thus, skip connections help the model efficiently learn the residual of the desired output with respect to its inputs.