NeuralNetworks project RT(RNN + LSTM) solution needed. Write a function to detecte lead-lag relationship between two timeseries (vector<double>)

Hello, I need a function written in c++ (qt, ubuntu) that takes input 2 vector<double> and outputs which one of the 2 vectors is behind (lagging) and by how much. The solution has to implement reinforcement-learning (RT) techniques (Q-Learning?) on a RNN (recurent neural net) using your choice library (propose, explain why the choice). The NeuralNetwork will be "default" trained on some data i will provide in a file and the function you will write is expected to keep on training itself and get better at finding by how much the lagging vector is behind, as more data arrives. The code has to run fast, c++ ONLY is expected. possible declaration: struct Out { string Who; double HowMuch; }; Out LeadLag(const vector<double>& A, const vector<double>& B) {} A and B vectors contain the same amount of data all the time and the values are correlated and cointegrated but can move in totally different ranges. Example1 (picture example): A = { 1000, 1001, 1001, 1001, 998} B= { 999, 999, 1002 , 997, 997} Example2, totally different ranges but same relative movement: A = { 1000, 1001, 1001, 1001, 998} B = {0.05, 0.05, 0.08, 0.03, 0.03} More about the functionality: The vectors will be continously updated when new data arrives, both at the same time. (A.push_back(new_value), B.push_back(new_value) outside the function. Then, the function, will be called every time for each update of the data. The function has to return who of the two vectors is behind (A or B) and by how much X. For example1 (picture). At T1, the function has to return({"B",1}). since B is latent by 1. At T3 function will return({"A",3}) since A is latent by 3. *Please bid only if you have good experience with neural netowrks and c++. *Do not bid until you've read the requirement fully and you think you can do this. *Payment will be done 100% only at the end of the project.A Partial solution is not a solution. Code has to run and detect the pattern with at least 80% accuracy. p.s You can propose other solutions that are not about neural netowrks. In the end the main problem is to detect who lags and by how much out of the two vectors.