Multiobjective Simulation

This repository provides tools for running multi-objective service selection experiments. It includes several algorithms and helpers for analysing the resulting error and cost trade-offs.

Example visualisation

An example script demonstrates how to run a small experiment and plot the results using matplotlib.

Requirements

pip install matplotlib

Running the example

From the repository root execute:

python examples/plot_results.py

The script runs run_experiment with a tiny configuration and displays error and cost trends for two algorithms alongside their error–cost tradeoff across time.

Plotting CLI results

To visualise the output from a CLI run saved to results.json, first execute:

python -m multiobjective.cli run --out results.json

Then load the JSON and generate plots:

python - <<'PY'
import json
from multiobjective.plotting import plot_metric_over_time, plot_tradeoff, plot_indicator_metric

# Load CLI output
with open("results.json") as f:
    results = json.load(f)

series = results["series"]
times = range(len(next(iter(series.values()))["errors"]["tp"]))
algs = list(series)

# Plot trajectories
plot_metric_over_time(times,
    [series[a]["errors"]["tp"] for a in algs],
    algs, "Topology error over time", "Normalised error")
plot_metric_over_time(times,
    [series[a]["costs"]["tp"] for a in algs],
    algs, "Cost over time", "Normalised cost")

# Plot a quality indicator (e.g. hypervolume)
plot_indicator_metric(
    results["indicators"],
    "HV",
    "tp",
    "Topology hypervolume over time",
    "Hypervolume",
)

# Plot trade-off over time
plot_tradeoff(
    [series[a]["errors"]["tp"] for a in algs],
    [series[a]["costs"]["tp"] for a in algs],
    algs, "Error–cost tradeoff over time")
PY

Plotting indicator metrics

The experiment also logs common quality indicators such as hypervolume (HV), inverted generational distance (IGD) and additive epsilon (EPS) for each algorithm over time. The helper function indicator_series from multiobjective.plotting extracts these values in a convenient form for visualisation with plot_metric_over_time.

python - <<'PY'
import json
from multiobjective.plotting import indicator_series, plot_metric_over_time

with open("results.json") as f:
    results = json.load(f)

# Collect HV values for topology error across all algorithms
inds = results["indicators"]
times, hv_series, algs = indicator_series(inds, "HV", err_type="tp")

plot_metric_over_time(times, hv_series, algs,
                      "Hypervolume over time", "HV")
PY

The resulting plot shows how the hypervolume indicator evolves for each algorithm, enabling direct comparison of their performance.