| import os |
|
|
| import matplotlib.pyplot as plt |
| import networkx as nx |
| import numpy as np |
| import pandas as pd |
| import yaml |
|
|
|
|
| def create_topology(num_agents, topology_type, seed=None): |
| if topology_type == 'complete': |
| return nx.complete_graph(num_agents) |
|
|
| if topology_type == 'line': |
| return nx.path_graph(num_agents) |
|
|
| if topology_type == 'ring': |
| return nx.cycle_graph(num_agents) |
|
|
| if topology_type == 'grid': |
| n_rows = int(np.sqrt(num_agents)) |
| n_cols = -(-num_agents // n_rows) |
| G = nx.grid_2d_graph(n_rows, n_cols) |
| G = nx.convert_node_labels_to_integers(G) |
| if len(G) > num_agents: |
| G = nx.Graph(G.subgraph(range(num_agents))) |
| return G |
|
|
| if topology_type == 'random': |
| if num_agents <= 1: |
| G = nx.Graph() |
| G.add_node(0) |
| return G |
| p = max(2 * np.log(num_agents) / num_agents, 0.3) |
| rng_seed = seed if seed is not None else 0 |
| for _ in range(10): |
| G = nx.erdos_renyi_graph(num_agents, min(p, 1.0), seed=rng_seed) |
| if nx.is_connected(G): |
| return G |
| p *= 1.5 |
| return nx.complete_graph(num_agents) |
|
|
| raise ValueError(f"Unknown topology type: {topology_type!r}") |
|
|
|
|
| def plot_convergence(results_df, output_dir, filename="convergence.png"): |
| plt.figure(figsize=(12, 8)) |
| groups = results_df.groupby(['optimizer', 'function', 'topology', 'num_agents', 'dimensions']) |
| for (opt, func, topo, n_agents, dim), group in groups: |
| mean_obj = group.groupby('iteration')['best_global_objective'].mean() |
| std_obj = group.groupby('iteration')['best_global_objective'].std().fillna(0) |
| label = f'{opt} - {func} ({topo}, N={n_agents}, D={dim})' |
| plt.plot(mean_obj.index, mean_obj.values, label=label) |
| plt.fill_between(mean_obj.index, |
| mean_obj.values - std_obj.values, |
| mean_obj.values + std_obj.values, |
| alpha=0.15) |
|
|
| plt.xlabel('Iteration') |
| plt.ylabel('Best Objective Value') |
| plt.title('Optimization Convergence Comparison') |
| plt.yscale('log') |
| plt.legend(fontsize=7) |
| plt.grid(True) |
| plt.tight_layout() |
| os.makedirs(output_dir, exist_ok=True) |
| plt.savefig(os.path.join(output_dir, filename)) |
| plt.close() |
| print(f"Plot saved to {os.path.join(output_dir, filename)}") |
|
|
|
|
| def save_results_to_csv(results_df, output_dir, filename="experiment_results.csv"): |
| os.makedirs(output_dir, exist_ok=True) |
| filepath = os.path.join(output_dir, filename) |
| results_df.to_csv(filepath, index=False) |
| print(f"Results saved to {filepath}") |
|
|
|
|
| def load_config(config_path='config.yaml'): |
| with open(config_path, 'r') as f: |
| return yaml.safe_load(f) |
|
|
|
|
| def print_summary(results_df): |
| summary = results_df.groupby(['function', 'optimizer'])['best_global_objective'].min().unstack() |
| print("\n--- Experiment Summary (best value reached) ---") |
| print(summary.to_string()) |
| print() |
| for func in summary.index: |
| best_opt = summary.loc[func].idxmin() |
| best_val = summary.loc[func].min() |
| marker = "*** QIMAD WINS ***" if best_opt == 'QIMAD' else f"Winner: {best_opt}" |
| print(f" {func}: {marker} ({best_val:.4e})") |
|
|
