Sum-Rate Maximization in OFDMA Downlink Systems

I have attached a paper that aims at allocating radio resources: MCS index, Resource Blocks, and Power for the users of just one base station. The authors modeled the problem as a MILP problem, but they didn’t use a MILP solver. Instead, they developed their own algorithm. We are only interested in the MILP problem.
here is a link of the matlab documentation for MILP :
https://fr.mathworks.com/help/optim/ug/intlinprog.html

you have to solve The MILP problem of the paper in MATLAB
Solving the MILP problem in the paper using MATLAB, GUROBI, or CPLEX; solvers that are faster than MATLAB and can be called from within Matlab.
Then you will run the simulation for multiple times; you will registers the inputs: For eg. Position of users, noise channel gain, throughput demand…., and you will register the output: Users allocation, MCS indexes, RBs, Power on RBs
You will use the inputs and outputs to train a deep learning model( maybe a recurrent neural network) using deep learning toolbox in Matlab
Then compare the performance of the RNN model to the ILP model on a data set that wasn’t used for training.