The general concept of regression -- it can be linear or nonlinear and have one or more continuous or discrete outputs. It's good to understand that a multi-lay

The general concept of regression -- it can be linear or nonlinear and have one or more continuous or discrete outputs. It's good to understand that a multi-layer neural network and a simple linear regression are similar in some ways. Related to regression and gradient descent (which you mentioned) is maximum likelihood estimation. Some knowledge of a command line, whether it's bash on Linux or even Windows CMD, can aid productivity.