ML/AI Interview Prep Guide

coding/ml_coding_challenges.py#!/usr/bin/env python3 """ ML Coding Challenges -- From-Scratch Implementations ==================================================== Five classic machine-learning building blocks implemented in pure Python using ONLY the standard library (no NumPy, no scikit-learn). These are the kinds of "implement X from scratch" problems that come up in ML engineering interviews. Implementations: 1. train_test_split -- reproducible data splitting 2. KNNClassifier -- k-nearest-neighbors classification 3. KMeans -- Lloyd's algorithm for clustering 4. LinearRegressionGD -- linear regression via gradient descent 5. classification metrics -- accuracy, precision, recall, F1 Run it: python3 ml_coding_challenges.py Every section has a self-test that prints expected vs. actual and asserts correctness, so a clean run proves all implementations are correct. Requires: Python 3.10+ standard library only. """ from __future__ import annotations import math import random from collections import Counter # --------------------------------------------------------------------------- # 1. Train/test split # --------------------------------------------------------------------------- def train_test_split(X, y, test_size=0.2, seed=None): """Split parallel lists X and y into train/test partitions. Returns (X_train, X_test, y_train, y_test). The split is reproducible when a seed is provided. Shuffling is done on indices so X and y stay aligned.