AI & Machine Learning Puzzle Book

Most books about artificial intelligence explain what the algorithms do. This one hands you the problem first and lets you work it out. Which door leads to freedom when one guard always lies? Which fruit is this, given only its weight and size? Did this model actually learn the rule — or did it just memorise the answers? Solve each puzzle by hand, then solve it again in Python, using the exact reasoning a real machine-learning algorithm uses. By the end you will have built working versions of k-nearest neighbors, the perceptron, Naive Bayes and linear regression from scratch — no libraries, no calculus, no prior programming experience required. Covers train/test splits, cross-validation, precision and recall, decision trees with entropy and Gini impurity, search algorithms including A* and Eight Queens, Bayes’ theorem with Monty Hall and the medical-test paradox, k-means clustering with the elbow method, and why one perceptron cannot solve XOR — followed by exactly what a hidden layer fixes. A capstone chapter builds and evaluates a complete AI system end to end, and a closing chapter connects every idea back to scikit-learn. Every line of code was run and checked against its printed output before publication.