A coordinated effort to formalize the mathematics of operations research in Lean 4.
Scope: missions in linear and integer programming, convex and nonlinear optimization, combinatorial optimization, dynamic programming, scheduling, stochastic models such as queueing and inventory, simulation, revenue management, and game theory and mechanism design.
A Modern Introduction to Online Learning XI: Constrained Online Convex Optimization Reduces to Unconstrained Learning on Distance-Penalized Surrogate LossesTextbook
An online learner repeatedly chooses a point, observes a loss, and is judged by the difference between its accumulated loss and that of a fixed competitor. In practice, an available learner may operate on a broad feasible set W, while the application permits decisions only in a smaller set V. Replacing each output by a feasible point changes the learner's…
A Modern Introduction to Online Learning XIII: Online Mirror Descent Has Dynamic Regret (B_ψ(u_T; x_1) + Q·P(u_1, …, u_T))/η + (η/(2λ)) Σ_t ‖g_t‖⋆²Textbook
In online learning, a learner chooses a point before seeing each round's loss. Ordinary regret compares the learner with one fixed point chosen after all rounds. That comparison can be too coarse when a useful action changes over time. A portfolio allocation, a prediction rule, or an operating point may need to move after the environment shifts. Dynamic…
Fast Rates for Contextual Linear Optimization 3: Under the Margin Condition, a Plug-In Policy Whose Estimate Has Pointwise Sub-Gaussian Tails at Rate aₙ Has Regret O(B·aₙ^(−(1+α)/2))Research Paper
Many operational decisions are linear programs whose cost vector is uncertain but predictable from side information: routing with travel times that depend on weather and time of day, procurement with prices that depend on market signals, assignment with costs that depend on observed features. Contextual linear optimization (CLO) formalizes this: observe…
A Differential Equation for Modeling Nesterov's Accelerated Gradient Method: Theory and Insights V: The Speed Restarted ODE Converges Linearly for Strongly Convex fResearch Paper
Nesterov's accelerated gradient method minimizes a convex function f with L-Lipschitz gradient at the rate O(1/k2), which is optimal among first-order methods. For a μ-strongly convex f the method with the generic momentum coefficient (k−1)/(k+2) does not exploit the strong convexity: Su, Boyd and Candès show in §4.2 of arXiv:1503.01243 that its rate…
Phase Recovery, MaxCut and Complex Semidefinite Programming 1: For Injective A, Nonzero b and Solvable |Ax| = b, PhaseCutMod Is Tight Whenever PhaseLift IsResearch Paper
Phase recovery (also called phase retrieval) asks to reconstruct a complex signal from the magnitudes of linear measurements of it, the phases being lost. It arises wherever detectors record intensities only: X-ray crystallography, diffraction imaging, astronomy and optics, where a sensor measures ∣⟨ai,x⟩∣2 and not ⟨ai,x⟩. The…
Relatively Smooth Convex Optimization by First-Order Methods, and Applications III: The Primal Gradient Scheme Solves D-Optimal Design to Accuracy ε within 2n ln(2(f(x⁰)−f*)/ε)/ε IterationsResearch Paper
In D-optimal design, a statistician chooses weights for a collection of candidate observations. The weights determine an information matrix, and maximizing its determinant favors a design that estimates several parameters jointly with a small confidence region. The same determinant objective appears in the minimum-volume enclosing ellipsoid problem. These…
A Modern Introduction to Online Learning V: Online Mirror Descent with a λ-Strongly Convex Regularizer Has Regret max_t B_ψ(u; x_t)/η_T + Σ_t η_t‖g_t‖⋆²/(2λ)Textbook
Online learning studies repeated decisions made before their losses are known. A learner chooses a point, observes a loss function, and then uses that observation to choose the next point. The natural performance measure is regret: the learner's cumulative loss minus the cumulative loss of one fixed competitor chosen after the sequence is known. A regret…
Robust Markov Decision Processes 2: The Affine Reward-to-Go Semidefinite Program Is Exact for One Quadratic Constraint and Conservative for SeveralResearch Paper
A Markov decision process (MDP) evaluates a policy by its expected discounted reward, which requires the transition probabilities. In practice these are estimated from data, and the optimal policy of the estimated model can perform badly under the true kernel. Robust MDPs replace the kernel by an ambiguity set of kernels and judge a policy by its worst-case…
Distributionally Robust Convex Optimization II: The Naïve Approximation Implies the Infimal Convolution Bound, Which Implies the Distributionally Robust ConstraintResearch Paper
A constraint E[v(x,z~)]≤w on a decision x needs the distribution of the random vector z~, which in practice is known only through partial information: a support, some moments, the probability mass of a few regions. Distributionally robust optimization requires the constraint for every distribution consistent with that information.…
Robustness May Be at Odds with Accuracy 1: Every Classifier with Standard Accuracy ≥ 1 − δ Has ℓ∞ Robust Accuracy ≤ pδ/(1 − p) Against ε ≥ 2ηResearch Paper
Classifiers trained to be robust against small, adversarially chosen input perturbations routinely lose standard accuracy. Adversarial training (Madry et al., 2018) improves worst-case accuracy under an ℓp-bounded attack, but on MNIST, CIFAR-10 and ImageNet the resulting models are measurably less accurate on unperturbed inputs, and the loss grows with…
Limit Theorems for Nearly Unstable Hawkes Processes 3: The Nearly Unstable Hawkes-Based Price Model Converges in Law to a Heston-Type ProcessResearch Paper
High-frequency prices move by discrete ticks, and successive moves are strongly correlated: an upward tick tends to be followed by a downward one. Bacry, Delattre, Hoffmann and Muzy (BDHM12, BDHM13b) proposed to model the upward and downward price jumps as the two components of a bidimensional Hawkes process, a pair of counting processes in which every…
An Introduction to Matrix Concentration Inequalities II: A Random Row-and-Column Submatrix Z = PBR Satisfies E‖Z‖² ≤ 3(p/d)(r/n)‖B‖² Plus Logarithmic Row, Column and Entry TermsTextbook
Large matrices are often reduced by retaining only some rows or columns. The reduced matrix is cheaper to store and apply, but its spectral norm can be distorted by a single unusually large row, column, or entry. A useful estimate must show both the average effect of sampling and the cost of these local concentrations. In §5.2 of Tropp's monograph, matrix…
Intrinsic Robustness of the Price of Anarchy II: Outcome Sequences with Vanishing Average Regret Have Average Cost at Most the Robust Price of Anarchy Times OptimalResearch Paper
In a game, each participant chooses a strategy while the cost of that choice depends on what everyone else does. A network user may choose a route, for example, but its travel time depends on the routes chosen by other users. The price of anarchy compares the cost of self-interested outcomes with the best joint outcome. A guarantee restricted to a Nash…
A Differential Equation for Modeling Nesterov's Accelerated Gradient Method: Theory and Insights IV: For Strongly Convex f, f(X(t)) − f⋆ ≤ C‖x₀ − x⋆‖²/(μ^((α−2)/2) t^α) for 2 ≤ α ≤ 2r/3Research Paper
Accelerated gradient methods minimize smooth convex functions with a characteristic momentum term. Su, Boyd, and Candès identified a continuous-time differential equation whose trajectories model that acceleration. The equation replaces an iteration index by continuous time and makes the effects of damping visible. This mission concerns the paper's strongly…
Relatively Smooth Convex Optimization by First-Order Methods, and Applications II: The Dual Averaging Scheme Attains min_{i≤k} f(x^i) − f(x) ≤ μh(x)/((1+μ/(L−μ))^k − 1) ≤ (L−μ)h(x)/kResearch Paper
Classical complexity bounds for first-order methods assume that the gradient of the objective is Lipschitz continuous, or that the objective is strongly convex, with respect to a norm. Many convex problems of practical interest satisfy neither assumption: the objective of D-optimal experimental design, −lndet(∑ixiaiai⊤), has a gradient that…
An Introduction to Matrix Concentration Inequalities I: A Matrix Gaussian Series Satisfies v(Z) ≤ E‖Z‖² ≤ 2v(Z)(1 + log(d₁ + d₂))Textbook
Random matrices arise when observations, signs, or model coefficients fluctuate. Their spectral norm measures the largest amplification they can cause to a vector. A bound on that norm can turn entrywise randomness into a statement about the behavior of an entire linear operator. Tropp's chapter develops a common framework for Gaussian and Rademacher sums…
A Differential Equation for Modeling Nesterov's Accelerated Gradient Method: Theory and Insights I: The ODE Ẍ + (3/t)Ẋ + ∇f(X) = 0 with X(0) = x₀, Ẋ(0) = 0 Has a Unique Global SolutionResearch Paper
Nesterov’s accelerated gradient method is a discrete procedure for minimizing a smooth convex function. Its iterates use both a gradient step and a momentum term, so the position at the next step depends on recent motion as well as on the current gradient. Su, Boyd, and Candès study a continuous-time model of that behavior: a trajectory whose acceleration…
A Modern Introduction to Online Learning XV: Online Learning Bounds Rademacher Complexity, Giving Massart's Lemma G∞√(2 ln d)/√TTextbook
Learning theory asks whether a rule that fits an observed sample will also perform well on future observations. A class with many functions can fit accidental patterns in the sample, so its size and geometry affect the reliability of empirical risk estimates. Empirical Rademacher complexity measures this ability to fit noise: attach an independent random…
A Modern Introduction to Online Learning X: The Weighted Average Algorithm with a Uniform Prior Has Regret (d/α)(1 + ln(T/d + 1)) on α-Exp-Concave LossesTextbook
Online prediction compares the losses of a sequence of decisions with the losses of one fixed decision chosen after the sequence is known. A learner would like a regret guarantee that remains valid for every such competitor, even when the losses arrive in an arbitrary order. Exponentially weighted averaging is a way to make each decision from all feasible…
Isotropic Gaussian Random Fields on the Sphere: Regularity, Fast Simulation and Stochastic Partial Differential Equations I: u ∈ V^η(−1,1) iff Σ u_ℓ²(2ℓ+1)(1+ℓ^{2η}) < ∞Research Paper
A centered isotropic Gaussian random fieldT on the unit sphere S2 is determined by its angular power spectrum(Aℓ)ℓ≥0: its covariance is
A Modern Introduction to Online Learning XIV: Optimistic FTRL Self-Play on a Smooth Convex-Concave Function Has Duality Gap at Most a Constant over TTextbook
A saddle-point problem asks one player to choose a decision that keeps a payoff small while another chooses a decision that makes it large. Such problems include two-player zero-sum games and constrained optimization written through a Lagrangian. Online convex optimization supplies a way to construct approximate saddle points from repeated play: each player…
A Modern Introduction to Online Learning VI: Follow-the-Regularized-Leader with a µ-Strongly Convex Regularizer Has Regret ((ψ(u) − min ψ)/α + α)·L√T/√µTextbook
In online convex optimization, a learner chooses a point before seeing the loss assigned to that round. Its performance is measured against a fixed point that could have been chosen after the full sequence became known. A learner that merely minimizes accumulated past losses can react sharply when successive losses favor different points. Follow-the-Regulari…
A Modern Introduction to Online Learning II: Online-to-Batch Conversion Turns a Regret Bound into a Risk Bound with Deviation 2√(2 ln(2/δ)/T)Textbook
An online learning algorithm chooses a decision before seeing each round's data and is judged by regret: how much more loss it incurs than a fixed decision chosen in hindsight. Statistical learning asks a different question. After seeing data drawn from a distribution, how small is the decision's expected loss on another draw? Chapter 3 of Orabona's textbook…
Linear Convergence of Gradient and Proximal-Gradient Methods Under the Polyak-Łojasiewicz Condition III: Randomized Coordinate Descent Has Expected Rate (1 − μ/(dL))^kResearch Paper
Coordinate descent changes one component of a parameter vector at a time. This matters when evaluating or updating a single component is cheaper than handling the full vector, for example in problems with sparse features. The coordinate chosen at each step can be random or selected from the current gradient. Karimi, Nutini, and Schmidt study these choices…
Joint Assortment Optimization and Customization under a Mixture of Multinomial Logit Models 2: Customization Raises the Optimal Expected Revenue by at Most a Factor m, and the Factor Is TightResearch Paper
An assortment decision determines which products a seller makes available to a customer. When customers differ, the seller may first choose a limited collection of products and then decide which of those products to display to each arriving customer. This is a concrete way to represent personalized product pages or offers with a common inventory or display…