Machine Learning · 19 problems · 8 optional · 3 project ideas

Problem sheet

Beginner

8 problems

Features, Labels and Baseline Models

supervised-learning · features · labels

The scikit-learn Estimator API

scikit-learn · estimator-api · regression-metrics

Regression MetricsOptional

metrics · mse · mae

Splitting Data with scikit-learn

scikit-learn · train-test-split · stratification

Seeded Train/Test SplitOptional

train-test-split · stratification · generalisation

Confusion Matrix, Precision, Recall, F1

metrics · confusion-matrix · precision

Scaling inside a Pipeline

scikit-learn · pipeline · standardisation

Standard Scaler (fit on train only)Optional

preprocessing · standardization · leakage

Intermediate

9 problems

Least-Squares Line

linear-regression · least-squares · closed-form

Linear Regression by Gradient Descent

gradient-descent · linear-regression · mse

Logistic Regression from Scratch

logistic-regression · log-loss · sigmoid

k-Nearest Neighbours with scikit-learn

scikit-learn · knn · hyperparameters

k-Nearest Neighbors Classifier by HandOptional

knn · instance-based-learning · distance

Polynomial Features and Validation Curves

scikit-learn · overfitting · model-complexity

Polynomial Features and Overfitting by HandOptional

polynomial-regression · overfitting · underfitting

Cross-Validation with scikit-learn

scikit-learn · cross-validation · kfold

Pipelines and ColumnTransformer

scikit-learn · pipeline · column-transformer

Advanced

10 problems

k-Fold Cross-Validation by HandOptional

cross-validation · model-selection · folds

Regularisation and Hyperparameter Search

scikit-learn · regularisation · ridge

Ridge Regression (Closed Form)Optional

regularisation · ridge · l2

Decision Stump with Gini Impurity

decision-trees · gini · splitting

Decision Trees and Random Forests

scikit-learn · decision-trees · random-forest

Gradient Boosting

scikit-learn · gradient-boosting · ensembles

Thresholds, ROC and Precision–Recall Curves

scikit-learn · roc-auc · precision-recall

k-Means and Silhouette Scores

scikit-learn · clustering · kmeans

k-Means Clustering by HandOptional

clustering · kmeans · unsupervised-learning

Capstone: an End-to-End Tabular Model

scikit-learn · end-to-end · pipeline

Project ideas

3 project ideas

House-Price Predictor

Project idea · regression · pipelines · feature-engineering

Spam or Fraud Classifier

Project idea · classification · imbalanced-data · thresholds

Serve Your Model with FastAPI

Project idea · deployment · fastapi · model-serving