Handbook of Metaheuristic Algorithms
Authors: Chun-Wei Tsai & Ming-Chao Chiang
Publisher: Academic Press
The required textbook provides an overview of optimization and heuristic models and has a number of metaheuristic models. We will only go through a few in class, but this book may serve as a good resource in the future.
Metaheuristic algorithms can be considered the epitome of unsupervised learning algorithms for the optimization of engineering and artificial intelligence problems.
Examples include:
Distinct from most supervised learning algorithms that need labeled data to learn and construct determination models, metaheuristic algorithms inherit characteristics of unsupervised learning algorithms used for solving complex engineering optimization problems without labeled data, just like self-learning, to find solutions to complex problems.
Although readers may be able to find source code for some metaheuristic algorithms on the Internet, the coding styles and explanations are generally quite different, and thus requiring expanded knowledge between theory and implementation.
We will not cover the dept of knowledge in the book, and instead, just get a taste for metaheuristics. We will focus on process and how to adapt and make the right decisions given the problem type.
We will also review Optimization, which is a lead into Heuristic Modelling.
The goal of optimization is to find the best solution among many possible options within a solution space. The process generally involves:
Find good solutions; estimate these solutions; and finally use the information thus obtained to make a decision. OR
Simply announce that the solution for the problem in question has been found.
Thinking Through Optimization: What’s one part of your daily routine that you’d love to optimize? If you could use a heuristic to make it easier or faster, how would it work?
Given a set of cities and the distance between each pair of cities, find the shortest path to visit each city and then return to the city of departure.
To calculate how many paths there are in a TSP problem, an author might divide (n−1)! by 2 to account for the fact that in the Traveling Salesman Problem (TSP), routes that are the reverse of each other are considered equivalent.
In this example, r1-r3 are unique routes and r4-r6 are direct reversals of r1-r3.
This formula, (n−1)! /2 is commonly used when considering the TSP under the assumption that the direction of travel does not matter, which is often the case. 
Different paths for a truck driver in the case of four cities.
import numpy as np
from scipy.special import factorial
# TSP paths calculation
np.set_printoptions(suppress=True)
n = np.arange(4, 12)
result = factorial(n - 1) / 2
print(result) # Expected output: [3, 12, 60, 360, 2520, 20160, 181440, 1814400][ 3. 12. 60. 360. 2520. 20160. 181440. 1814400.]
How to Write and Read and Algorithm
Summary of Optimization
Travelling Salesman
Greedy Algorithm
Some Benchmark Problems
Exhaustive Search
Hill Climbing
Metaheuristics
Algorithm -> Pseudocode -> Implementation in Python 
Besides summarizing Optimization and Heuristic Models, we will cover Simulated Annealing and Genetic Algorithms. Some example of Metaheuristics are below:
Local Search-Based Metaheuristics
Evolutionary Computation
Swarm Intelligence
Human Intelligence
Nature-Inspired Algorithms: Inspired by various natural and biological phenomena.