Skip to content

Tutorials

Learn a workflow from its first run to an interpretable result. Start with the route that matches your question; each link below leads to an existing worked example. For a single operation, use the How-to Guides.

Choose a learning path

Use the search field for a topic such as constraints, an algorithm such as MOEA/D, or a public API name such as StudySpec. On a narrow screen, the magnifying-glass button opens search and the menu button opens the sections. Each numbered lesson provides its Python download, setup, and previous/next steps. Wide tables scroll horizontally; focus a table to scroll with arrow keys.

Start with the three fundamental lessons. Each has complete downloadable code, actual output and a figure, an interpretation, and one guided change. They are verified against the published VAMOS 1.0.1 wheel; use the shared setup for their exact dependency recipe.

Lesson Outcome
T01. Your first meaningful result Run ZDT1 and distinguish the population from its non-dominated subset.
T02. Solve your own constrained problem Validate the evaluator, check feasibility, and preserve row alignment.
T03. From a Pareto set to a decision State scales and preferences, select a feasible alternative, and recover its decisions.

Continue with a complete experiment. The study setup provides three standalone scripts, a small functional run, and a larger recorded example. The recorded report exposes every measured row.

Lesson Outcome
T06. Run a fair comparative study Fix the matrix, seeds, budget and incident policy before execution.
T07. Measure quality and progress with budget Measure saved fronts using fixed references and compare fresh runs at increasing budgets.
T08. Turn study results into a report Rebuild tables and figures from retained observations and qualify the conclusions.
Your goal Follow this route What you will understand
Try VAMOS Set up → T01 → Result reference guide How to run a built-in problem, relate decisions to objectives, and distinguish a population from its non-dominated subset.
Solve your own problem T02 → T03 → Full modeling guide How to express bounds and requirements, check feasibility, and select a candidate using explicit preferences.
Run a reproducible study T06: plan and run → T07: measure → T08: report How to make the design explicit and trace each table, figure and conclusion to preserved evidence.

New to the Python workflow? The Minimal Python Track provides a command-line entry point. Prefer executable files? The Examples gallery has a small script for each of the three routes, with setup and source links.

Learn an algorithm

After a first run, choose a guide by the mechanism you want to understand. Each guide includes runnable code, an actual result, and its limits.

Guide Focus Before you start
NSGA-II Non-dominated sorting, diversity, and explicit configuration on ZDT1. Read a two-objective result.
MOEA/D Decomposition and the role of weight vectors and neighborhoods on ZDT1. Understand objective trade-offs and evaluation budgets.
NSGA-III Reference directions and population size on three-objective DTLZ2. Understand a Pareto front and why more objectives change the problem.
  • SMS-EMOA: explicit configuration, recorded results and mechanism-specific limits.
  • SPEA2: explicit configuration, recorded results and mechanism-specific limits.
  • IBEA: explicit configuration, recorded results and mechanism-specific limits.
  • SMPSO: explicit configuration, recorded results and mechanism-specific limits.
  • AGE-MOEA: explicit configuration, recorded results and mechanism-specific limits.
  • RVEA: explicit configuration, recorded results and mechanism-specific limits.

Use the algorithm capability matrix to check encoding, constraint, archive, and backend support before choosing a family. The configuration reference gives the exact fields for all built-in algorithms.

Continue with a specific task

The small study demonstrates the workflow; it does not establish an algorithm ranking. Preserve that distinction when moving from a tutorial to your own experiment. Check the stability policy and the dependencies stated by each destination before using optional features.

Representations, applications and advanced experiments

The coverage setup supplies the common verified environment and evidence.

Goal Worked lesson
Preserve binary, integer, permutation or mixed decisions Decision representations
Select a feasible mixed design Engineering design
Trade prediction error against selected features Feature selection
Search model settings under a complexity proxy Model hyperparameters
Separate algorithm-training and validation seeds Experimental tuning
Remove one search component under a fixed design Controlled ablation
Measure initialization and subsequent execution Backend performance