APLENDE/

marketplace /

Classical Machine Learning in Depth for Tabular and Unsupervised Work

by cinojosa0705 · Classical ML algorithms in depth · taken 1×

A 10-hour, mastery-oriented course on decision trees, random forests, gradient boosting, SVMs, k-means, PCA, t-SNE, and UMAP, with from-scratch implementations and a Kaggle-style capstone. Designed from the supplied learner profile and curriculum constraints.

Free account, then it's yours: your own private copy with a clean slate and its own review schedule.

Sign up free to take it →

4

modules

12

lessons

11h

estimated work · target 10h

What's inside

52 tasks · 48 flashcards · 73 concepts · 37 recall prompts

remember
6
understand
12
apply
9
analyze
17
evaluate
6
create
2

Sample recall prompts — this is what studying it feels like

  • From memory, define Gini impurity and explain how a decision tree converts node impurities into split gain.
  • Explain why an unrestricted decision tree can fit training data extremely well yet generalize poorly, and name three controls that reduce this problem.
  • Contrast the decision boundary and feature assumptions of a classification tree with logistic regression.

Syllabus

Trees from First Principles

Build decision trees and random forests from scratch, then reason about their bias, variance, and diagnostics.

  • Classification Trees: Splits, Impurity, and Stopping flashcard · quiz · project · reading · 8 cards
  • Regression Trees and Recursive Implementation note · quiz · project · reading
  • Random Forests: Bagging, Decorrelation, and Out-of-Bag Evaluation reading · flashcard · quiz · project · teach_back · 8 cards

Boosted Trees for Tabular Prediction

Understand boosting as stagewise loss minimization, then use XGBoost and LightGBM within a leakage-resistant model-selection workflow.

  • Gradient Boosting as Functional Gradient Descent flashcard · quiz · project · reading · 8 cards
  • XGBoost: Regularized Second-Order Tree Boosting quiz · note · project · reading
  • LightGBM and a Disciplined Tabular Workflow reading · flashcard · quiz · project · essay · 8 cards

Margins and Unsupervised Structure

Develop geometric and practical working knowledge of SVMs, then learn what k-means optimizes and how to diagnose misleading clusters.

  • Support Vector Machines: Margins, Hinge Loss, and Kernels flashcard · quiz · project · reading · 8 cards
  • SVM Practice: Scaling, Hyperparameters, and Multiclass Decisions note · quiz · project · reading
  • k-Means: Objective, Lloyd's Algorithm, and Diagnostics note · quiz · project · teach_back · reading

Dimensionality Reduction and Integrated Practice

Derive PCA from covariance eigendecomposition, use t-SNE and UMAP responsibly, and complete an end-to-end tabular capstone.

  • PCA from Covariance Eigendecomposition flashcard · quiz · project · reading · 8 cards
  • t-SNE and UMAP as Responsible Visualization Tools reading · quiz · project · teach_back
  • Kaggle-Style Tabular Capstone essay · reading · note · quiz · project

Concepts it teaches

Recursive PartitioningNode ImpuritySplit GainTree Stopping RulesLeaf Class ProbabilityRegression TreeSquared-Error ImpurityLeaf Mean OptimalityRecursive Tree BuilderCost-Complexity PruningBaggingBootstrap SamplingRandom Feature SubsamplingEnsemble VarianceOut-of-Bag EvaluationPermutation ImportanceAdditive Tree ModelGradient BoostingPseudo-ResidualShrinkageWeak LearnerXGBoost ObjectiveSecond-Order Loss ApproximationGradient and Hessian StatisticsRegularized Leaf WeightXGBoost Split GainEarly StoppingHistogram BinningLeaf-Wise Tree GrowthLeaf-Count CapacityNative Categorical SplitsTabular Validation DesignData LeakageMaximum-Margin HyperplaneSupport VectorHinge LossSoft-Margin SVMKernel TrickRBF KernelFeature ScalingSVM C ParameterSVM Gamma ParameterMulticlass ReductionProbability CalibrationSVM Computational Costk-Means ObjectiveLloyd's Algorithmk-means++ InitializationInertiaSilhouette ScoreCluster StabilityDistance-Based ScalingPCA CenteringCovariance MatrixEigendecompositionPrincipal ComponentExplained Variance RatioPCA ProjectionPCA Reconstructiont-SNEt-SNE PerplexityEmbedding KL DivergenceUMAP Neighbor GraphUMAP n_neighbors and min_distEmbedding TrustworthinessEmbedding StabilityEnd-to-End Tabular WorkflowBaseline ModelOut-of-Fold PredictionControlled Hyperparameter SearchModel ComparisonError Slice AnalysisReproducible ML Experiment