Advanced Topics
What Lives Here
The advanced track is where the site moves into:
- real analysis
- optimization
- numerical methods
- ODEs and dynamical systems
- signal processing and estimation
- control and dynamics
- stochastic processes
- stochastic control and dynamic programming
- matrix analysis
- learning theory
- high-dimensional probability
- high-dimensional statistics
- information theory
Live Advanced Modules
Real Analysis
Real Analysis is the rigor bridge from calculus into theorem-heavy work.
Its first-pass spine is:
- Rigorous Convergence
- Continuity, Compactness, and Completeness
- Sequences and Series of Functions
- Differentiation and Integration as Theorems
- Fixed-Point, Implicit, and Inverse Function Ideas
This module is where the site makes limits, continuity, compactness, and theorem-level calculus explicit enough for advanced probability, optimization, and learning theory.
Optimization
Optimization is the first fully mature advanced module with a complete five-page spine:
- Convex Sets and Separation
- Convex Functions and Subgradients
- Constrained Optimization, KKT, and Lagrangians
- Unconstrained First-Order Methods
- Duality and Certificates
It is the natural bridge from the foundation stack into convexity, duality, certificates, first-order methods, and research-facing mathematical modeling.
Numerical Methods
Numerical Methods is the computation bridge from exact mathematics to finite-precision algorithms.
Its first-pass route is:
- Floating-Point, Conditioning, and Backward Error
- Numerical Linear Systems and Factorizations
- Iterative Methods and Preconditioning
- Numerical Least Squares and Regularization
- Eigenvalue and SVD Computation
- Approximation, Differentiation, Integration, and Error Control
- Time-Stepping for ODEs and Stability - optional extension toward ODEs and simulation
This module is where the site turns exact linear-algebra, calculus, and optimization objects into actual computations that live in floating point and must be judged by conditioning, stability, and error.
Signal Processing and Estimation
Signal Processing and Estimation is the signal-and-systems bridge from mathematical functions and sequences into filtering, sensing, communication, and inverse problems.
Its first-pass spine is:
- Signals, Convolution, and Linear Time-Invariant Systems
- Fourier Analysis, Frequency Response, and Spectral Views
- Sampling, Aliasing, and Reconstruction
- Noise Models, Wiener Filtering, and MMSE Estimation
- State Estimation, Smoothing, and Hidden-State Inference
- Inverse Problems, Deconvolution, and Regularized Recovery
- Signal Processing Bridges to Communication, Sensing, and Modern ML
This module is where the site turns ordered data, system response, convolution, sampling, noise models, hidden-state inference, and ill-posed recovery into a full bridge toward communication, sensing, and modern ML.
ODEs and Dynamical Systems
ODEs and Dynamical Systems is the continuous-time bridge from local rules for change to trajectories, equilibria, and stability.
Its first-pass route is:
- First-Order ODEs, Existence, and Solution Curves
- Second-Order Systems, State Variables, and Reduction to First Order
- Linear Systems, Matrix Exponentials, and Modes
- Phase Portraits, Equilibria, and Local Stability
- Lyapunov Functions, Invariant Sets, and Long-Time Behavior
- Discretization, Time-Stepping, and the Bridge to Control
- Time-Stepping for ODEs and Stability - adjacent numerical bridge
- Research Bridges: Reverse-Time SDEs, Probability-Flow ODEs, Flow Matching, and Control - optional research bridge into continuous-time ML
This module is where the site turns derivatives into actual continuous-time models, qualitative dynamics, and the language needed for later control and simulation.
Control and Dynamics
Control and Dynamics is the systems-facing bridge from passive dynamics to feedback, estimation, and steering.
Its first-pass spine is:
- State-Space Models, Inputs, and Outputs
- Controllability, Reachability, and Observability
- Feedback, Stability, and Pole Placement
- Linear Quadratic Regulation and Riccati Intuition
- Estimation, Kalman Filtering, and the Separation Principle
- Model Predictive Control and Constraint Handling
- Learning-Based Control, System Identification, and RL Bridges
Closest companion pages:
This module is where the site turns trajectories into systems with actuation, sensing, feedback, and optimization over behavior.
Stochastic Control and Dynamic Programming
Stochastic Control and Dynamic Programming is the uncertainty-aware bridge from control and probability into sequential decision-making.
Its first-pass spine is:
- Controlled Markov Models, Policies, and Cost Functionals
- Finite-Horizon Dynamic Programming and Backward Induction
- Infinite-Horizon Value Functions, Bellman Equations, and Contractions
- Value Iteration, Policy Iteration, and Approximate Dynamic Programming
- Stochastic Linear Systems, LQG, and the Separation Principle
- Continuous-Time Stochastic Control and Hamilton-Jacobi-Bellman Intuition
- Partial Observability, Belief States, and RL/Control Bridges
This module is where the site turns sequential choices under uncertainty into explicit objects like policies, transition laws, cost-to-go structure, belief states, and planning under hidden information.
Matrix Analysis
Matrix Analysis is the operator-level bridge between linear algebra and modern theory-facing math.
Its first-pass spine is:
- Norms and Operator Norms
- Positive Semidefinite Matrices and Quadratic Forms
- Spectral Inequalities and Variational Principles
- Perturbation and Stability
- Trace, Determinant, and Matrix Functions
This module is where the site turns matrices into operator-sized, spectral, quadratic-form, and matrix-function objects instead of only arrays with eigenvalues.
Learning Theory
Learning Theory is the bridge from the site’s math backbone into theorem-level ML guarantees.
Its first-pass spine is:
- ERM, Population Risk, and Hypothesis Classes
- PAC Learning, Sample Complexity, and the Learning Setup
- VC Dimension and Shattering
- Uniform Convergence and Generalization Bounds
- Rademacher Complexity and Data-Dependent Capacity
- Algorithmic Stability and Regularization
- Generalization in Modern Regimes
This module is where the site turns empirical risk, population risk, hypothesis classes, and generalization into explicit mathematical objects.
High-Dimensional Probability
High-Dimensional Probability is the next theory-facing layer after classical probability and learning theory.
Its first-pass route is:
- Concentration Beyond Basics
- Sub-Gaussian and Sub-Exponential Variables
- Random Vectors, Isotropy, and Norms
- Random Matrices and Spectral Concentration
- High-Dimensional Phenomena
- High-Dimensional Probability for Learning Theory and Modern ML
This module is where the site shifts from scalar probability intuition to non-asymptotic control of maxima, norms, random vectors, and random matrices.
High-Dimensional Statistics
High-Dimensional Statistics is the statistical layer that sits directly on top of high-dimensional probability, matrix analysis, and optimization.
Its first-pass route is:
- Sparsity and Regularization
- Lasso and Compressed Sensing Basics
- Design Geometry: Restricted Eigenvalues, Coherence, and RIP
- High-Dimensional Regression
- Covariance, PCA, and Spectral Estimation in High Dimension
- Minimax and Lower Bounds
- Inference in High Dimension
This module is where the site turns p >> n, sparse structure, shrinkage, and recoverability into explicit statistical objects instead of vague intuitions about “many features.”
Information Theory
Information Theory is the uncertainty-and-limits bridge connecting probability and statistics to coding, communication, and many ML objectives.
Its first-pass route is:
- Entropy, Cross-Entropy, and KL Divergence
- Mutual Information, Conditional Entropy, and Data Processing
- Typicality, Source Coding, and Compression Intuition
- Channel Coding, Capacity, and Converse Proofs
- Rate-Distortion and Representation Tradeoffs
- Variational Objectives, ELBO, and Information Bounds
- Information-Theoretic Lower Bounds in Statistics, Learning, and Communication
This module is where the site turns entropy, divergence, coding cost, and communication limits into explicit reusable objects instead of isolated formulas.
Stochastic Processes
Stochastic Processes is the probability-over-time bridge from ordinary random variables into chains, martingales, diffusions, and long-run stochastic behavior.
Its first-pass route is:
- Markov Chains and Stationary Distributions
- Martingales and Optional Stopping Intuition
- Poisson Processes and Counting Models
- Brownian Motion and Diffusion Intuition
- SDEs and Ito Intuition
- Mixing, Ergodicity, and MCMC Bridges
This module is where the site turns one-shot probabilistic intuition into random evolution over time and the first long-run objects that matter in control, sampling, and diffusion-side modeling.