to accelerate Greedy Heuristic search even more

you can use templates

bound by lower AND upper bound

Take Greedy algorithm as a base.
For random fit: sort items randomly.
For First fit: dont sort the items.

Requires extending GetSortedItemIds() method in class Problem.

Implement true iterative Brute Force.
Name the fuction IterativeBruteForce.
Rename old brute force to some BranchAndBound

and then do template specializations

this should NOT save problem files to disc. Generate one problem, solve it, (optionally save solution), repeat.

Add pages for each solving function.
Lay out theory: sources, complexity, result, error margin.
Paste code.
Provide example.
Explain arguments.

Implement true DFS Brute force.
It should first reach full depth and then check value of the solution it came up with.
For non full depth calls it should return better solution of all sub calls.
Name function DFSBruteForce
Rename old Brute force to some BranchAndBound

Solve one problem in each variant with every algorithm and check if for example brute force gives better results than greedy or if solutions for path variant are better than for cycle variant.

int main(int argc, char *argv[]) { ... }

• argument for generate: -file [file]

It should chceck if found cycle fits in the knapsack

Implement Floyd Algorithm.

count every visit

ad -structure flag that wolud override structure to find from problem file

Options:

• imgui
• python matplotlib

MiniZinc: https://www.minizinc.org/
Implement a function that would convert Problem class object to MniZinc notation

Specifically in method Consume()

In Batch Solve output additional .json file with average solve time and quality.
And average found max_value and remaining spaces.

Implement Greedyfied version of Floyd Approximation Algorithm.

that uses topo sort

Implement Ford-Bellman Algorithm.

Add method for solving continuous knapsack problem.
And use it as a bounding method in branch and bound algorithm.