Radio resource allocation in communication networks is essential to achieve optimal performance and resource utilization. In modern interference networks the corresponding optimization problems are often nonconvex and their solution requires significant computational resources. Hence, practical systems usually use algorithms with no or only weak optimality guarantees for complexity reasons. Nevertheless, asserting the quality of these methods requires the knowledge of the globally optimal solution. State-of-the-art global optimization approaches mostly employ Tuy's monotonic optimization framework which has some major drawbacks, especially when dealing with fractional objectives or complicated feasible sets. In this thesis, two novel global optimization frameworks are developed. The first is based on the successive incumbent transcending (SIT) scheme to avoid numerical problems with complicated feasible sets. It inherently differentiates between convex and nonconvex variables, preserving the low computational complexity in the number of convex variables without the need for cumbersome decomposition methods. It also treats fractional objectives directly without the need of Dinkelbach's algorithm. Benchmarks show that it is several orders of magnitude faster than state-of-the-art algorithms. The second optimization framework is named mixed monotonic programming (MMP) and generalizes monotonic optimization. At its core is a novel bounding mechanism accompanied by an efficient BB implementation that helps exploit partial monotonicity without requiring a reformulation in terms of difference of increasing (DI) functions. While this often leads to better bounds and faster convergence, the main benefit is its versatility. Numerical experiments show that MMP can outperform monotonic programming by a few orders of magnitude, both in run time and memory consumption. Both frameworks are applied to maximize throughput and energy efficiency (EE) in wireless interference networks. In the first application scenario, MMP is applied to evaluate the EE gain rate splitting might provide over point-to-point codes in Gaussian interference channels. In the second scenario, the SIT based algorithm is applied to study throughput and EE for multi-way relay channels with amplify-and-forward relaying. In both cases, rate splitting gains of up to 4.5% are observed, even though some limiting assumptions have been made.