Free course · by Debugshala, Indore
Machine Learning Tutorial
27 free lessons, in the order we teach them in class: start at lesson 1 and work down. Every lesson has an analogy, a diagram, code you can run or copy, a quick quiz and an FAQ.
- 27Lessons
- 4Modules
- ~5 hReading time
- FreeAlways
01Machine Learning Foundations
- AI vs ML vs Deep Learning: Differences and Types of LearningAI vs ML vs deep learning explained with a nested-circles diagram, plus supervised, unsupervised and reinforcement learning labs you can run in Python.
- Overfitting and the Bias-Variance Trade-off, with a NumPy DemoUnderstand overfitting, underfitting and the bias-variance trade-off by fitting polynomials in NumPy, then fix it with more data, ridge and cross-validation.
- Model Evaluation Metrics: Accuracy, F1, ROC-AUC, RMSE and MoreModel evaluation metrics with runnable sklearn code: train/test split, confusion matrix, precision, recall, F1, ROC-AUC, MAE, RMSE, R² and when to use each.
- Cross-Validation Explained: K-Fold, Stratified and Time SeriesCross-validation explained with runnable scikit-learn code: k-fold, stratified, time series and group splits, leakage-free pipelines and how to read the scores.
- Ridge, Lasso and Elastic Net: Regularization in PythonRegularization in machine learning made clear: how ridge shrinks, lasso zeroes and Elastic Net blends coefficients, why scaling matters and how to pick alpha.
- Hyperparameter Tuning: Grid, Random and Bayesian SearchHyperparameter tuning in scikit-learn: GridSearchCV vs RandomizedSearchCV, a hands-on Bayesian optimisation loop, Optuna, and a practical tuning workflow.
- Feature Selection in Machine Learning: Filter, Wrapper, EmbeddedFeature selection methods in Python: variance and correlation filters, mutual information, RFECV, lasso and permutation importance, plus the leakage trap.
- Handling Imbalanced Data: Class Weights, SMOTE and ThresholdsHandle imbalanced data in machine learning: why accuracy fails, class weights, oversampling, undersampling, SMOTE, threshold tuning and PR-AUC, with code.
02Supervised Learning
- Linear Regression Explained: Least Squares to scikit-learnLinear regression explained with rickshaw fares and rent data: least squares, coefficients, gradient descent, residual checks and assumptions, in Python.
- Logistic Regression Explained: Sigmoid, Odds and Python CodeLogistic regression explained: how the sigmoid turns a score into a probability, odds ratios, log loss, softmax for many classes and the C parameter in sklearn.
- K-Nearest Neighbours (KNN) Algorithm Explained with PythonK-nearest neighbours explained: how KNN votes, why scaling matters, how to choose k with cross-validation, KNN regression and the curse of dimensionality.
- Naive Bayes Classifier Explained: Bayes Rule to Spam FilterNaive Bayes classifier explained with a WhatsApp spam filter: priors, likelihoods, Laplace smoothing, Gaussian vs Multinomial NB and scikit-learn code.
- Support Vector Machines (SVM) Explained: Margins and KernelsSupport vector machines explained: maximum margin, support vectors, the C parameter, the kernel trick, RBF gamma and when SVMs beat other models, in Python.
- Decision Trees Explained: Gini, Entropy and scikit-learn LabDecision trees explained with a delivery example: compute Gini and entropy by hand, find the best split, then train DecisionTreeClassifier and print its rules.
- Random Forest Explained: Bagging, OOB Score and ImportanceRandom forest explained: bootstrap samples, random feature splits, the out-of-bag score, how many trees you need, feature importance traps and limits.
- Gradient Boosting and XGBoost Explained with Python CodeGradient boosting explained step by step: residual-fitting trees, learning rate, early stopping, HistGradientBoosting, and what XGBoost adds on top.
- Ensemble Learning: Bagging, Boosting and Stacking in PythonEnsemble learning explained: why combining models works, voting, bagging, boosting and stacking with out-of-fold predictions, plus when not to use one.
03Unsupervised Learning
- K-Means Clustering Explained: Algorithm, Elbow and SilhouetteK-means clustering explained with dark-store planning: the assign and update loop, scaling, choosing k with elbow and silhouette, k-means++ and limits.
- Hierarchical Clustering Explained: Dendrograms and LinkageHierarchical clustering explained: how agglomerative merging works, reading a dendrogram, single vs complete vs average vs Ward linkage, and cutting the tree.
- DBSCAN Clustering Explained: eps, min_samples and NoiseDBSCAN clustering explained: core, border and noise points, choosing eps with a k-distance plot, DBSCAN vs k-means on curved data, and when to use HDBSCAN.
- PCA in Machine Learning: Dimensionality Reduction in PythonPCA explained step by step: covariance, eigenvectors, explained variance, choosing components, scaling and pipelines, with runnable numpy and sklearn code.
- Anomaly Detection in Machine Learning: Isolation Forest, LOFAnomaly detection techniques in Python: Isolation Forest, Local Outlier Factor, One-Class SVM, novelty detection, thresholds and how to evaluate detectors.
04Advanced ML
- Recommender Systems Explained: Collaborative Filtering in PythonHow recommender systems work: content-based and collaborative filtering, matrix factorisation, hit rate evaluation and cold start, with runnable Python code.
- Transfer Learning Explained: Feature Extraction vs Fine-TuningTransfer learning explained: why pretrained layers transfer, feature extraction vs fine-tuning, when to use each, plus PyTorch and Hugging Face examples.
- Federated Learning Explained: FedAvg, Privacy and a Python DemoFederated learning explained: how FedAvg trains one model without moving data, non-IID clients, secure aggregation and real uses, with numpy simulations.
- Machine Learning Projects for Beginners (with a Portfolio Plan)10 machine learning projects for beginners with dataset types, skills and difficulty, plus a GitHub README template and LinkedIn plan to land interviews.
- Machine Learning Interview Questions: 30 Answers with Code30 machine learning interview questions with crisp answers and runnable code: bias-variance, leakage, regularisation, metrics, algorithms and scenarios.