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.
Optimality of the Lookahead Peak-Shaving Policy for Robust Inventory with Volume FlexibilityResearch Paper
Many manufacturers buy transport or production capacity under contracts with volume flexibility: in every period a base capacitykn can be sourced at the contracted unit price c, and any volume above it is still available but costs an extra flexibility premiumcp per unit. The sourcing cost per period is therefore piecewise linear and convex, with slopes c…
Markov Decision Processes with Applications to Finance X: Terminal Wealth under Partial Observation Is Solved by a Bellman Equation in Wealth and FilterTextbook
The distribution of stock returns is not constant: drifts and volatilities change with economic conditions that are not directly observed, and drift parameters in particular are notoriously hard to estimate. A standard way to model this in discrete time is to let the return distribution depend on an unobservable Markov chain, the hidden state or regime,…
Blind Network Revenue Management and Bandits with Knapsacks Under Limited Switches: LS-2SLP and a Matching Lower Bound Pin the Optimal Regret at T^(1/(2-2^(-ν(s,d))))Research Paper
Network revenue management studies a firm that sells several products made from shared, limited resources over a finite season: airline seats on connected legs, hotel room nights, retail items that share inventory. In the blind version the firm does not know how demand responds to price and has to learn it while selling. Besbes and Zeevi (2012) posed this…
A New Approach for Vehicle Routing with Stochastic Demand: Combining Route Assignment with Process Flexibility: Adjacent Overlapping Costs Q·r∞/(N′µ) Times Reoptimization as the Fleet GrowsResearch Paper
A distributor that serves the same customers every day with a fleet of capacitated vehicles faces a vehicle routing problem with stochastic demand: the customers are fixed, but their daily demands are random. Re-solving the vehicle routing problem every morning (reoptimization) gives the cheapest routes, but it is expensive to compute and changes every…
Dynamic Resource Allocation: Algorithmic Design Principles and Spectrum of Achievable Performances: Every Online Multisecretary Policy Has Regret Ω(T^{1/2−1/(2(1+β))}) on a Distribution With a GapResearch Paper
Online resource allocation asks a decision maker to spend a fixed budget on requests that arrive one at a time, each decision irrevocable. Network revenue management, online matching and order fulfillment are instances. The multisecretary problem is the one-resource case: T candidates with abilities drawn i.i.d. from a known distribution F on [0,1] arrive…
When Variability Trumps Volatility: Optimal Control and Value of Reverse Logistics in Supply Chains with Multiple Flows of Product: Without Demand Variability, Reverse Orders Stop After Period N−j+1Research Paper
Firms increasingly move product upstream as well as downstream: excess stock at a retail or regional location is returned to a warehouse, resold to a distributor, or held for a later period. Practitioner and academic writing on reverse logistics usually credits it with value when demand is volatile (Jaaron and Backhouse, 2016; Lieckens and Vandaele, 2007,…
An Online Learning Approach to Dynamic Pricing and Capacity Sizing in Service Systems: GOLiQ Has Gradient Bias at Most K0/(8k) and Regret O((logML)2) in a GI/GI/1 QueueResearch Paper
A service provider who runs a single-server queue chooses two numbers: the fee p charged to each customer, which determines the demand rate λ(p), and the service capacity μ, which costs c(μ) per unit time. Waiting customers cost h0 per unit time. When the distributions of interarrival and service times are known, the profit-maximizing pair …
An Asymptotically Tight Learning Algorithm for Mobile-Promotion Platforms: The Policy BidAlloc Has Regret at Most K2K12TTI Over I SeasonsResearch Paper
Mobile-promotion platforms sell advertisers a number of ad impressions to be delivered within a time window, and buy those impressions one at a time by bidding on an ad exchange. A bid that is too low misses a campaign's target, which costs a penalty; a bid that is too high overpays. The difficulty is that the platform does not know its win curves, the…
The Power of Simple Menus in Robust Selling Mechanisms: Robust Randomized Pricing over Fixed Price Levels Is a Finite Linear ProgramResearch Paper
A seller offers one item to one buyer whose valuation lies in [0,vˉ] and whose valuation distribution F is not known. Robust mechanism design asks for a selling mechanism that performs well against every distribution consistent with the seller's partial information. Without restrictions on the menu, the optimal robust mechanism randomizes over a…
Optimal Learning for Structured Bandits: DUSA Has Asymptotic Regret at Most (1+ε)(C(P)+ε(1+Σₓ∆(x,P))) log T, Matching the Lower Bound C(P) as ε → 0, with O(log T) Exploration RoundsResearch Paper
In a multi-armed bandit a learner repeatedly pulls one of finitely many arms and observes a random reward, trying to earn as much as an omniscient learner who knows the best arm. In many operations problems the arms are not unrelated: in dynamic pricing the probability of no sale is monotone in the price, in assortment or dosage problems mean rewards vary…
Technical Note—On Revenue Management with Strategic Customers Choosing When and What to Buy: Every Dynamic Pricing Policy Earns at Most the Fluid Value Jˉ1Research Paper
Network revenue management studies a seller of perishable capacity: flight legs, hotel nights, cloud resources. Products are bundles of resources, the inventory is fixed, and the season ends at a known time. Upper bounds on optimal revenue are the basic tool of the field, because an exact optimum is out of reach and every performance guarantee for a…
Convolution Bounds on Quantile Aggregation: The Worst-Case Quantile of a Sum with Given Marginals Is at Most the Convolution Bound, with Equality in Five CasesResearch Paper
A bank or an insurer that holds n risks X1,…,Xn usually has good statistical information about each risk separately and very little about how they depend on each other. Regulatory capital is computed from a quantile of the aggregate loss X1+⋯+Xn (the Value-at-Risk, VaR), so the question "how large can the quantile of the sum be, over all…
Convergence and Stability of Coupled Belief–Strategy Learning Dynamics in Continuous Games: Beliefs and Strategies Converge Almost Surely to a Payoff-Consistent Equilibrium Fixed PointResearch Paper
Many strategic systems learn two things at once. Travellers on a road network, firms in a market or bidders in an auction adjust their strategies to each other, and an information platform (a navigation service, a market maker, a public agency) estimates an unknown parameter of the environment from the payoffs those strategies produce. The payoffs are…
Optimal Oracle Inequalities for Projected Fixed-Point Equations, with Applications to Policy Evaluation: Averaged Stochastic Approximation Attains the Instance-Dependent Approximation FactorResearch Paper
Many estimation problems ask for the solution v∗ of a linear fixed-point equation v=Lv+b on a high- or infinite-dimensional space, when only noisy unbiased samples of L and b are available. The leading example is policy evaluation in reinforcement learning: the value function of a Markov reward process solves the Bellman equation v=r+γPv, and…
Coordinate Descent Without Coordinates: Tangent Subspace Descent on Riemannian Manifolds: Under a (γ, r)-Gap-Ensuring Rule, Gradients Vanish, Limit Points Are Stationary, G-Convex Gaps Are O(1/t)Research Paper
Block coordinate descent (BCD) minimizes a function on Rn by taking a gradient step in one block of coordinates at a time. It is a standard method for large problems, and its convergence theory for smooth and convex objectives is well developed (Beck & Tetruashvili, 2013). Many optimization problems, however, are posed on a Riemannian manifold:…
Column-Randomized Linear Programs: Performance Guarantees and Applications: with probability ≥ 1 − δ, Δv(P_J) ≤ Δv(P_distr) + C(1 + mγ‖A‖max)(1 + √(2 log(2/δ)))/√KResearch Paper
Many linear programs in operations research have a small number m of constraints and an astronomically large number n of columns: cutting-stock patterns, rankings in nonparametric choice estimation, state–action pairs of a Markov decision process. The standard remedy is column generation, which needs a problem-specific pricing subproblem. Akchen and Mišić (a…
Dynamic Development Contests: Probabilistic Change-of-State Disclosure Minimizes the Expected Lead Time of a Budget-Constrained ContestResearch Paper
Governments and firms often buy development work through contests: suppliers race through milestones, and the first to finish gets a prize. Such contests stall when a supplier learns that it has fallen behind and stops working, so the organizer's control over what suppliers learn is a design lever alongside the prize. Khorasani, Körpeoğlu and Krishnan study…
Multiplayer Bandits Without Observing Collision Information: A Decentralized Algorithm Has Expected Regret O(mK log(T)/Δ²)Research Paper
In cognitive radio networks, several radios share a set of frequency channels. Each radio picks a channel in every time slot. The quality of a channel is random and unknown, and two radios that pick the same channel interfere and transmit nothing. This is the multiplayer multi-armed bandit problem: m players repeatedly choose among K arms, an arm pays a…
Strategy-Proof Multi-Issue Mediation: An Application to Online Dispute Resolution: A Mediation Rule Is Efficient, Individually Rational and Strategy-Proof Iff It Is an Adjacent RuleResearch Paper
Mediation is a dispute resolution method in which a neutral third party helps disputants reach a voluntary agreement short of litigation. Courts in all US states offer some form of it, and online platforms use automated mechanisms to resolve large numbers of small disputes (Kesten and Özyurt, 2018, p.2). A mediator who proposes an outcome from the parties'…
Bayesian and Randomized Clock Auctions: A Deterministic Single-Price Clock Auction Achieves an O(loglogk)-Approximation to the Expected Optimal WelfareResearch Paper
A seller offers a service to a set of buyers, each with a private value, and a feasibility constraint limits which sets of buyers can be served together. This is the basic single-parameter problem of mechanism design. Milgrom and Segal proposed clock auctions for it, and designed one for reallocating radio frequency licenses (Milgrom and Segal 2020). In a…
Bilateral Trade: A Regret Minimization Perspective: Under Realistic Feedback and Independent Bounded-Density Valuations, Scouting Bandits Has Regret O(MT^{2/3}) via an Explicit Bound in T₀, K and αResearch Paper
In bilateral trade a seller and a buyer each hold a private valuation for one good, and an intermediary (a platform or broker) decides at which price the good changes hands. Myerson and Satterthwaite showed that no mechanism can be at once efficient, incentive compatible, individually rational and budget balanced (Myerson–Satterthwaite 1983). Among…
Mean Field Contest with Singularity: the stopping contest has a unique mean field equilibrium, given in closed formResearch Paper
Rank-based rewards are the standard incentive in sales tournaments, research prizes and promotion contests. Seel and Strack (Gambling in contests, J. Econ. Theory 2013) showed that such contests can be analysed as games of risk taking, in which each player chooses when to stop a privately observed diffusion. Nutz and Zhang (Reward design in risk-taking…
Preservation of Additive Convexity and Its Applications in Stochastic Optimization Problems: optimal prioritized transformation preserves additive convexity and bounded sensitivityResearch Paper
Additive convexity is the structural property behind the tractability of serial inventory theory. Clark and Scarf (1960) showed that the multi-variate value functions of a serial inventory system are sums of univariate convex functions of the echelon inventory levels. This gives the optimality of echelon base-stock policies and replaces a high-dimensional…
Price Delegation with Learning Agents: a static separable contract implements the fluid first-best policy to within any ε under limited liability and limited inventoryResearch Paper
Many firms let their salespeople set prices. A salesperson who spends effort talking to a customer can learn what that customer is willing to pay and quote a price close to it; a salesperson who skips that effort can only quote a price that works on average. The firm sees neither the effort nor the customer's value, only the revenue of each sale, and it can…
On Linear Programming for Constrained and Unconstrained Average-Cost Markov Decision Processes with Countable Action Spaces and Strictly Unbounded Costs: No Duality Gap Under MajorizationResearch Paper
A Markov decision process (MDP) with the average-cost criterion asks for a policy that minimizes the long-run expected cost per stage. A classical way to solve it is linear programming: the optimal average cost is the value of a linear program over occupation measures (probability measures on state–action pairs that are invariant under the dynamics), and…