Dept. of Statistics & Insurance Science · University of Piraeus

Statistical
Machine Learning

A rigorous treatment of probabilistic models, optimization theory, and algorithmic decision-making — from convex loss surfaces to ensemble methods and deep architectures.

Course Instructor Ioannis E. Livieris Associate Professor
Course Overview
DepartmentStatistics & Ins. Sci.
ECTS Credits6.0
Supervised Models8 algorithms
Unsupervised Models3 algorithms
PrerequisitesLinear Algebra · Prob.
Semester2027 – 2028

Overview

What is Machine Learning?

Machine Learning allows systems to learn autonomously from data — identifying patterns, constructing models, and making decisions without being explicitly programmed for each task.

How machines learn

Just as humans build knowledge from experience, ML algorithms are exposed to data from which they extract features, relationships, and patterns. This acquired structure enables accurate predictions on new, unseen data — and continuous improvement as more data becomes available.

Why it matters

Human analysts face hard limits on the data they can process manually. ML systems operate at arbitrary scale, uncovering structure that would otherwise remain hidden. Applications span self-driving vehicles, fraud detection, medical diagnosis, and recommendation systems at Netflix, Amazon, and YouTube.

01 · Supervised
Supervised Learning

Trained on labeled data — each input is paired with the correct output. The model learns a mapping and generalizes to new examples. Covers both classification (discrete targets) and regression (continuous targets).

Random ForestLogistic RegressionSVMXGBoostANN
02 · Unsupervised
Unsupervised Learning

Works on unlabeled data, discovering hidden structures autonomously. No human labeling required, enabling the algorithm to scale to terabytes of raw data. Representations adapt dynamically post-deployment.

k-MeansPCADBSCANAssociation Rules
03 · Reinforcement
Reinforcement Learning

An agent interacts with an environment via a reward signal. Favorable actions are reinforced; unfavorable ones penalized. The algorithm iterates via trial-and-error until it discovers the optimal policy.

Q-LearningPolicy GradientActor-Critic

ML Models

Select any model to explore its theory, interactive visualizations, and real-world applications.

SupervisedRegression
Linear Regression

Estimates a continuous output by fitting the best-fit hyperplane, minimizing MSE. Fully interpretable — each β quantifies the exact linear contribution of a feature.

ŷ = β₀ + β₁x₁ + … + βₙxₙ
SupervisedClassification
Logistic Regression

Probabilistic binary classifier using the sigmoid function to map linear combinations to [0,1]. Decision boundary is a hyperplane; trained via maximum likelihood.

σ(z) = 1 / (1 + e−z)
SupervisedClassification
k-Nearest Neighbors

Non-parametric instance-based learner. Classifies by majority vote among k nearest neighbors. No training phase — all computation deferred to query time.

d(x,y) = √Σ(xᵢ−yᵢ)²
SupervisedClassification
Support Vector Machine

Finds the maximum-margin hyperplane separating classes. Kernel trick extends to non-linear boundaries. Dual formulation leads to sparse support vector solutions.

max 2/‖w‖ s.t. yᵢ(wᵀxᵢ+b)≥1
SupervisedClassification
Decision Tree

Recursive axis-aligned partitioning of feature space. Splits chosen to maximize information gain (ID3) or minimize Gini impurity (CART). Interpretable; prone to overfitting.

Gini = 1Σ pₖ²
VOTE
EnsembleClassification
Random Forest

Bagging ensemble of decision trees with random feature subsampling. Reduces variance via averaging. Provides feature importance scores. Robust to outliers and high-dimensional data.

ŷ = mode{T₁(x), …, TB(x)}
EnsembleBoosting
XGBoost

Gradient-boosted trees with L1/L2 regularization, column subsampling, and second-order Taylor expansion of the loss. Industry-standard for tabular data competitions.

Fm(x) = Fm−1(x) + η·hm(x)
SupervisedProbabilistic
Naïve Bayes

Probabilistic classifier based on Bayes' theorem with conditional independence assumption. Extremely fast; works well on text and high-dimensional sparse data.

P(y|x) ∝ P(y)ΠP(xᵢ|y)
Neural NetSupervised
Artificial Neural Network

Layered composition of parameterized non-linear transformations. Trained via backpropagation and gradient descent. Universal approximator; foundation of deep learning.

a(l) = σ(W(l)a(l−1) + b(l))
UnsupervisedClustering
k-Means Clustering

Iterative algorithm partitioning data into k clusters by alternating assignment and centroid update steps. Minimizes within-cluster sum of squared distances (inertia).

J = ΣΣ‖x − μₖ‖²
UnsupervisedDim. Reduction
Principal Component Analysis

Linear orthogonal projection that maximizes variance. Computes eigenvectors of the covariance matrix. Foundation of dimensionality reduction, noise filtering, and feature extraction.

C = XᵀX/(n−1), Cv = λv

Algorithm Landscape

Comparing the Models

Three perspectives on all 11 algorithms — their computational and explanatory footprint, how they cluster into families, and how much labeled data each one needs to learn effectively.

Scatter Plot
Complexity vs. Interpretability
Donut Chart
Distribution by Category
Bar Chart
Minimum Training Data Requirements

Theory

Mathematical Foundations

Core concepts from optimization, probability, and statistical learning theory that underpin every algorithm in the course.

📐
Optimization Theory

Gradient descent, SGD, momentum, and Adam. Convexity conditions, Lagrangian relaxation, and KKT optimality conditions for constrained problems.

GD / SGDAdamKKTConvexity
🎲
Probabilistic Models

Maximum Likelihood Estimation, MAP inference, Bayesian reasoning, conjugate priors, and posterior predictive distributions.

MLEMAPBayesianPriors
🎯
Generalization Theory

VC dimension, PAC learning framework, L1/L2 regularization, cross-validation strategies, and the bias-variance decomposition of generalization error.

VC Dim.PACL1/L2 Reg.CV
🔬
Kernel Methods

The kernel trick, Reproducing Kernel Hilbert Spaces, Mercer's theorem, and standard kernels: RBF, polynomial, and sigmoid.

RKHSRBFMercerGram Matrix
🌲
Ensemble Methods

Bagging and boosting theory: bias-variance reduction, AdaBoost convergence, gradient boosting as functional gradient descent.

BaggingBoostingAdaBoostGBDT
🔢
Matrix Decompositions

Eigendecomposition, SVD, and their application to dimensionality reduction, pseudo-inverses, and covariance geometry.

SVDEigenPCACovariance

Bias–Variance Tradeoff

Total expected error decomposes as Bias² + Variance + Irreducible Noise. Simple models underfit (high bias); complex models overfit (high variance). Optimal complexity sits at the minimum of total error.


Resources

Course Material

Textbooks, lecture notes, useful links, and the full semester syllabus.

📚 Bibliography
01
Pattern Recognition and Machine Learning
Bishop, C.M. (2006) · Springer
02
The Elements of Statistical Learning
Hastie, Tibshirani & Friedman (2009) · Springer
03
An Introduction to Statistical Learning
James, Witten, Hastie & Tibshirani (2021) · Springer
04
Understanding Machine Learning: From Theory to Algorithms
Shalev-Shwartz & Ben-David (2014) · Cambridge
🔗 Useful Links
scikit-learn Documentation
Python ML library — API reference & user guide
scikit-learn.org ↗
Kaggle Learn
Free micro-courses and competitions for practice
kaggle.com/learn ↗
StatQuest with Josh Starmer
Intuitive video explanations of ML algorithms
youtube.com/@statquest ↗
Distill.pub
Interactive, peer-reviewed ML research articles
distill.pub ↗
📄 Lecture Notes & Slides
W1
Introduction to Statistical Learning
PDF — upload coming soon
W2
Linear & Logistic Regression
PDF — upload coming soon
W3
Decision Trees & Ensemble Methods
PDF — upload coming soon
W4
Neural Networks & Backpropagation
PDF — upload coming soon
🗓 Semester Syllabus
  • Wk 1
    Introduction to ML
    Paradigms, applications, course structure
  • Wk 2
    Linear & Logistic Regression
    MLE, gradient descent, regularization
  • Wk 3
    k-NN & Naive Bayes
    Instance-based learning, probabilistic classifiers
  • Wk 4
    Support Vector Machines
    Maximum margin, kernel trick, dual problem
  • Wk 5
    Decision Trees & Random Forest
    CART, information gain, bagging
  • Wk 6
    Boosting & XGBoost
    AdaBoost, gradient boosting, regularization
  • Wk 7
    Neural Networks
    Backpropagation, activations, deep learning intro
  • Wk 8
    Unsupervised: k-Means & PCA
    Clustering, eigendecomposition, applications
  • Wk 9
    Model Evaluation & Selection
    CV, AUC-ROC, bias-variance, generalization
  • Wk 10
    Project Presentations
    Applied ML project · peer review