AdONE Seminar - Monday, December 1, 2025: Prof. Özlem Karsu (Bilkent University)

Finding fair solutions under uncertainty

We consider applications where decisions affect multiple beneficiaries, and the decision maker aims to ensure that the impact is distributed equitably among them. Our focus is on stochastic environments, where there is uncertainty in the system.

We first present a robust programming approach that seeks to maximize system efficiency while ensuring an equitable outcome allocation, even in the worst-case scenario. We derive tractable formulations by leveraging results on the properties of highly unfair allocations. We demonstrate the approach through applications in project selection and shelter allocation.

We then turn to chemotherapy scheduling and propose a metric for quantifying efficiency and fairness concerns in patient waiting times. To optimize this metric, we formulate a two-stage stochastic mixed-integer nonlinear programming model. We introduce a binary search-based algorithm that incorporates scenario set reduction and augmentation techniques to improve computational performance. We illustrate the applicability of the approach using data from a major oncology hospital.

Date: December 1, 2025

Time: 5 pm, s.t.

Place: Arcisstr. 21, room 0505.Z1.534Z/0505.Z1.536Z

 

AdONE Seminar - Monday, December 15, 2025: Prof. Claudia d'Ambrosio (École Polytechnique de Paris)

Statistical learning for solving complex mixed integer non linear optimization problems

Mixed Integer Non Linear Programming (MINLP) is a challenging class of Mathematical Programming (MP) problems where decision variables might be continuous or integer and the objective function and constraints are expressed as non linear functions of such variables. Modern general-purpose MINLP solvers shown impressive computational performance improvements on special MINLPs, for example when g(x) are convex or quadratic. However, when one (or more) function "complex", the MINLP can easily become practically intractable. A standard approach is to replace the complex functions with a surrogate approximation. Several different approaches have been proposed in the literature. Some of them are specific for some kind of non linear functions, for example univariate or bivariate ones. However, these methods cannot easily generalize to more general non linear functions. More recently, machine learning methods, which model could be represented as a mixed integer linear program. However, it is difficult to impose some properties of the function, when expert knowledge on those is available. To overcome such limitations, we propose a new method, based on Mixed-integer Smoothing

Surrogate Optimization with Constraints (MiSSOC). Our approach is composed of two steps: 1) Surrogate approximations are found by fitting a function that is the sum of univariate B-splines. The fitting is performed by solving a MP model where B-spline basis functions are fixed and their weight are found; 2) The resulting surrogate MINLP, obtained by replacing the complex functions with their surrogate approximations, is solved with a tailored solver called Sequential Convex-MINLP (SC-MINLP). We present promising computational results on some instances of the MINLPLib and a real-world application, the hydro unit commitment problem.

Date: December 15, 2025

Time: 5 pm, s.t.

Place: Arcisstr. 21, room 2760, HS (0507.02.760)

 

AdONE Seminar - Thursday, December 18, 2025: Prof. Dirk Bergemann (Yale University)

Information Design and Mechanism Design: An Integrated Framework

This survey develops an integrated framework for information design and mechanism design in screening environments. Using the language of majorization and quantile functions, we show that both problems reduce to maximizing linear functionals subject to majorization constraints. For mechanism design, the designer chooses allocations weakly majorized by the exogenous inventory. For information design, the designer chooses information structures that are mean-preserving contractions of the prior distribution. When the designer can choose both the mechanism and information structure simultaneously, the problem becomes bilinear with two majorization constraints. We show that pooling is always optimal in this case. Our approach unifies classical results in auction theory and screening, extends them to information design settings, and provides new insights into the welfare effects of joint optimization.

Date: December 18, 2025

Time: 4 pm, s.t.

Place: 00.04.011 HS 02, (5604.EG.011), MI, Campus Garching

 

AdONE Seminar - Monday, January 12, 2026: Prof. Rene Haijema (Wageningen University)

Dynamic expiration date-based discounting of fresh food products

To reduce food waste, many supermarkets discount food products that are close to their expiration date. In practice, this is done either by discount labels put on the product or by electronic shelf labels (or digital price tags) showing the price per expiration date. Digital price tags allow to easily change the price of products and to apply different discount rates to items with different expiration dates. An important question to practitioners is when and how much discount to offer. In this study, we use Stochastic Dynamic Programming (SDP) to derive optimal expiration-date-based discounting policies for a profit-maximizing retailer who sells a product with 𝑚 periods (e.g., days) of shelf life. We compare various discounting strategies, such as static last-day discounting, optimal dynamic last-day, and last-two-days discounting, against the no-discounting strategy. 

The model allows products of different expiration dates to be in stock simultaneously, as replenishment happens every period. In the last-day discounting policies, two selling prices co-exist: the regular price and the discounted price. When applying a last-two-days discounting policy, three selling prices co-exist. Demand and product withdrawal depend on both price and product age (freshness). We consider different customer picking behavior, and divide customers into First-Expiry-First-Out (FEFO) and Last-Expiry-First-Out (LEFO) consumers (i.e, customers that pick the oldest items first and customers that take the freshest items available). For LEFO customers, we also consider that a fraction of these customers will pick discounted old items (depending on the size of discount). Finally, extra demand is attracted as long as discounted products are available. 

Optimal policies are derived by SDP and evaluated by simulation to generate insights into the impact of discounting on profits, sales, fill rates, and waste. Various key factors, such as shelf life, customer picking behavior, and discount sensitivity are analyzed in detail. The results show that the last-two-days discounting policy performs well. Averaged over all experiments, this policy demonstrates a 3.8% increase in profits compared to no-discounting, and a waste reduction from 5.6% to 3.6%. Smaller, but still significant improvements are shown over simpler discounting policies
 

Date: January 12, 2026

Time: 5:15 pm

Place: Arcisstr. 21, room 0544, Seminarraum (0505.EG.544)

 

AdONE Seminar - Monday, January 19, 2026: Prof. Frits Spieksma (TU Eindhoven)

Title: tba

Abstract: tba

Date: January 19, 2026

Time: 5 pm, s.t.

Place: tba

 

AdONE Seminar - Monday, February 2, 2026: Maximilian von Aspern, Lukas Brandl, Felix Buld (TUM, AdONE)

Title: tba

Abstract: tba

Date: February 2, 2026

Time: 5 pm, s.t.

Place: Karlstr. 45, room 6009 (2906.DG.009)