Can some classic maximization algorithms be added?

[zzzkhj]

Currently, there is only one maximization algorithm, celf, in the code repository. Can the author add some other classic maximization algorithms? For example, IMM, degree discount, etc.

[eliotwrobson]

Sure, I have the code for these laying around somewhere actually. The code for most of these is pretty short, and not all package consumers will use it, so I'll probably just add the code to the documentation site.

[eliotwrobson]

I'm not sure about the IMM algorithm, but I have code for degree discount here:

def degree_discount_runner(
    graph: nx.DiGraph,
    budget: int,
) -> ExperimentResult:
    start_time = time.perf_counter()

    # Initialize data structures
    t_dict: defaultdict[int, int] = defaultdict(int)
    best_node_dict: DepqDict[int, float] = DepqDict()

    for node, out_degree in graph.out_degree():
        best_node_dict[node] = out_degree

    seeds = []
    times_taken = []

    for i in range(budget):
        node, _ = best_node_dict.pop_max_item()
        end_time = time.perf_counter()

        seeds.append(node)
        times_taken.append(end_time - start_time)

        if i == budget - 1:
            break

        for v in graph.neighbors(node):
            if v not in best_node_dict:
                continue

            t_dict[v] += 1

            d_v = graph.out_degree(v)
            t_v = t_dict[v]

            best_node_dict[v] = (
                d_v - 2 * t_v - (d_v - t_v) * t_v * graph[node][v]["activation_prob"]
            )

    assert len(seeds) == budget

This doesn't really use cynetdiff, so I probably won't add it to the repo.