August 28, 2021
Carnegie Mellon University
2021 YinzOR Student Conference
The CMU INFORMS Student Chapter is excited to host the 4th YinzOR Conference. This student conference
provides an opportunity for students whose research is in Operations Research, Management Science, Industrial Engineering, and other related fields to interact with each other.
This year, we are hosting the conference virtually over Zoom on August 28th!
Organizing Committee
CMU OR/OM/ChemE PhD Students
Tom Krumpolc (Co-Chair) - tkrumpol@andrew.cmu.edu
Mik Zlatin (Co-Chair) - mzlatin@andrew.cmu.edu
Violet Chen
Sagnik Das
Daniel de Roux
Anthony Karahalios
Neda Mirzaeian
Yanhan (Savannah) Tang
* Please feel free to email either co-chair with any of your questions/comments!
Past Conferences
YinzOR 2019
YinzOR 2018
YinzOR 2017
Video Submissions and Winners
We are happy to share the 6 great video submissions. Based on the voting results, we announce the 3 winners of the competition, congratulations!
We thank FedEx for sponsoring the prizes!
Conference Program
|
11:10am-12:00pm EST
|
Featured Talk: Optimizing over Trained Neural Networks: Neural Network Verification and Discrete Black-Box Optimization
|
|
Christian Tjandraatmadja, Google Research
|
|
We present applications and techniques for a special class of optimization problems where the objective function is the output of a neural network with fixed weights. These problems arise in two
types of applications: analyzing properties of neural networks and optimizing hard-to-model functions. In the first case, we are interested in measuring how robust a network is to adversarial attacks, that is, whether an adversary can change the input to manipulate a neural network into producing a result they desire. We show how a polyhedral approach to this problem can lead to efficient algorithms to certify that a neural network is robust against such attacks. In the second class of applications, the neural network plays the role of approximating a function that is difficult to model or optimize, possibly with complicating constraints. We present a framework for constrained discrete black-box optimization that uses neural networks as surrogate model and mixed-integer programming to select points to evaluate.
|
|
12:00pm-12:30pm EST
|
Labor Cost Free-Riding in the Gig Economy
|
|
Zhen Lian, Cornell GSM
|
|
We propose a theory of gig economies in which workers participate in a shared labor pool utilized by multiple firms. Since firms share the same pool of workers, they face a trade-off in setting pay rates; high pay rates are necessary to maintain a large worker pool and thus reduce the likelihood of lost demand, but they also lower a firm’s profit margin. We prove that larger firms pay more than smaller firms in the resulting pay equilibrium. These diseconomies of scale are strong too; firms smaller than a critical size pay the minimal rate possible (the workers’ reservation wage), while all firms larger than the critical size earn the same total profit regardless of size. This scale disadvantage in labor costs contradicts the conventional wisdom that gig companies enjoy strong network effects and suggests that small firms have significant incentives to join an existing gig economy, implying gig markets are highly contestable. Yet we also show that the formation of a gig economy requires the existence of a large firm, in the sense that an equilibrium without any firms participating only exists when no single firm has enough demand to form a gig economy on its own. The findings are consistent with stylized facts about the evolution of gig markets such as ride sharing.
|
|
12:30pm-1:00pm EST
|
Adaptive Clinical Trial Design with Surrogates: When Should We Bother?
|
|
Arielle Anderer, Wharton
|
|
Surrogate outcomes have long been used in clinical trials when the true outcome of interest is expensive, time consuming, or otherwise difficult to measure. In this work we propose optimal adaptive clinical trial designs that integrate surrogate and true outcomes, and we analytically and empirically characterize regimes where our designs are especially beneficial.
|
|
1:00pm-1:30pm EST
|
Lunch Break (Reminder to vote for the Video Contest!)
|
|
|
|
1:30pm-2:00pm EST
|
Social Pairings
|
|
|
|
2:00pm-2:50pm EST
|
Featured Talk: Data-driven Optimistic Optimization and Contextual Decision Making
|
|
Junyu Cao , McCombs, UT Austin
|
|
We study data-driven optimistic optimization (DOO)---the optimistic counterpart of data-driven robust optimization---as a systematic tool for online contextual decision-making. To capture the parameter uncertainty in contextual optimization, we propose a new uncertainty set which contains all parameters that satisfy the first-order optimality condition of a supervised learning problem with responses perturbed from some nominal values. Based on the proposed uncertainty set, we develop an algorithm which sequentially solves a max-max problem that seeks an optimal action. This builds a novel connection between DOO and upper-confidence-bound algorithms. We construct a computationally efficient approximation and establish its performance guarantees by deriving a nearly-optimal regret bound for a wide class of contextual decision-making problems.
|
|
2:50pm-3:20pm EST
|
Algorithmic Fairness in Suicide Prevention Interventions for the Homeless Youth Population
|
|
Aida Rahmattalabi CS, USC
|
|
Fueled by algorithmic advances, AI algorithms are increasingly being deployed in settings subject to unanticipated challenges
with complex social effects. Motivated by real-world deployment of AI driven, social-network based suicide prevention, we study the
robust graph covering problems subject to group fairness constraints. We show that, in the absence of fairness constraints,
state-of-the-art algorithms for the robust graph covering problem result in biased node coverage: they tend to discriminate
individuals (nodes) based on membership in traditionally marginalized groups. To mitigate this issue, we propose a novel formulation
of the robust graph covering problem with group fairness constraints and a tractable approximation scheme applicable to real-world
instances. We provide a formal analysis of the price of group fairness (PoF) for this problem, where we show that uncertainty can
lead to greater PoF. We demonstrate the effectiveness of our approach on several real-world social networks of youths experiencing homelessness.
Our method yields competitive node coverage while significantly improving group fairness relative to state-of-the-art methods.
|
|
3:20pm-3:50pm EST
|
Estimation of Customer Preferences for Attended Home Delivery
|
|
Pol Boada-Collado, Northwestern IEMS
|
|
Quantifying customers' sensibility to delivery lead time can allow retailers to balance customer satisfaction and shipping costs in attended home delivery. Using transactional data from a major furniture company committed to home delivery, we model how customers choose attended home delivery dates from among the set of alternative dates offered by the company. The model allows us to estimate the priorities that drive customer's choice and the relative impact of speed of delivery. The empirical results show that customers do not usually schedule attended home delivery for the days immediately after the purchase date. Speed of delivery is of limited importance relative to other priorities, such as day of week preferences or availability, when customers choose their preferred delivery date. For the company, this is an opportunity to slightly extend delivery windows and more efficiently utilize transportation capacity. Also, the empirical results show that customers are strongly segmented by day of the week of purchase, which can potentially be leveraged to decrease delivery lead times while increasing transportation capacity utilization. Our results show that companies committed to home delivery can more efficiently allocate costly delivery capacity while adapting last-mile operations to meet customer's expectations.
*Joint work with Sunil Chopra, Maria Ibanez, and Karen Smilowitz.
|
|
5:15pm-5:20pm EST
|
Video Contest winners will be announced!
|
