Lifting is defined as a map to a higher-dimensional space such that the solution of a mathematical problem in one space directly implies its solution in the other one. There are three major reasons for lifting a numerical problem into a higher dimensional space in the context of Newton-type solution methods: (a) enhancing the sparsity of linear systems within each Newton step; (b) improving the conditioning of linear systems within each Newton step; (c) reduction of nonlinearity for faster convergence, from one Newton step to the next. While the first two are important in some applications, they are often sacrificed in practice by a "condensing" procedure. The main reason for the success of lifted formulations in practice remains point (c), the reduction of nonlinearity, which is the one that is least understood so far. This is also the aspect "Lift to Learn" will focus on most strongly. Moreover, the application of lifting in the context of Deep Learning will be investigated.