This tutorial will focus on how the Machine Learning (ML) techniques applied to the evolving solution sets offered by Evolutionary Multi- and Many-objective Optimization Algorithms (EMaOAs). EMaOAs can facilitate knowledge discovery and performance enhancement across different phases of optimization, including, problem-modeling, optimal-search, and post-optimization decision-making. Beginning with the essential concepts in EMaO and ML domains, this tutorial will cover some representative studies endorsing the benefits of integrated use in these domains. It will highlight how ML intervention can facilitate: (i) better understanding of the problem-structure, (ii) dedicated operators to enhance the search efficacy of EMaOAs on convergence- and diversity-hard problems, and (iii) more efficient and customized decision-making. The presented approaches will be supported by exhaustive experimental results. Importantly, this tutorial will also emphasize the importance of not distorting the basic tenets of EMaOAs in pursuit of ML integration. A template to ensure the latter will be presented in light of ML-based risk-reward trade-off, exploration-exploitation balance, minimizing ad-hoc parameterization and avoiding extra solution evaluations. Notably, this tutorial will be interactive and will also include a walkthrough and execution of the python codes for the recently proposed ML-based operators to enhance EMaOA search efficacy. The tutorial will conclude with pointers to future directions in ML-assisted Evolutionary Multi- and Many-objective Optimization.

Syllabus:
1. Basic tenets of Evolutionary Multi- and Many-objective Optimization Algorithms (EMaOAs).
2. Basic tenets of Machine Learning (ML) techniques and synergy with EMaOAs.
3. Utility of ML intervention towards understanding the problem-structure better.
4. Dedicated ML-based operators for performance enhancement of EMaOAs, covering:
4a. Innovized operators (IP and IP2) for convergence enhancement (employing Artificial Neural Network (ANN) and Random Forests (RF)), with supporting results on convergence-hard test problems.
4b. Innovized operator (IP3) for diversity enhancement (employing k-Nearest Neighbors (kNN)), with supporting results on diversity-hard test problems.
4c. Unified Innovized operator (UIP) for simultaneous convergence and diversity enhancement (employing RF and kNN) with supporting results on both convergence and diversity-hard test problems.
4d. A template for Ml-assisted EMaAO: addressing the key considerations of ML-based risk-reward trade-off, exploration-exploitation balance, minimizing ad-hoc parameterization and avoiding extra solution evaluations
4e. A walkthrough and execution of codes implementing the above operators on test problems.
5. Utility of ML intervention towards more efficient and customized post-optimization decision-making.
6. Pointers to future directions in ML-assisted Evolutionary Multi- and Many-objective Optimization and brief discussion of EMaOA-assisted ML development.