Hybrid seminar announcement – Prof. Luciano Costa (UFPB), “Column generation applied to the estimation of non-parametric discrete-choice models”, May 26, 2025, 14:30
12 views
Skip to first unread message
Alice Raffaele
May 15, 2025, 3:40:24 AMMay 15
to airo-...@googlegroups.com, Luciano Costa
Dear AIROYoungers,
We hope you'll be able to join us either in person or remotely.
Best,
Alice
---------------------
Title: Column generation applied to the estimation of non-parametric discrete-choice models
Abstract: Discrete choice models (DCMs) provide probabilities for individuals choosing a certain alternative when faced with a set of limited options. DCMs can be parametric or non-parametric. Parametric models are easier to estimate but require assumptions about individuals' preferences, while non-parametric models rely solely on training data without any assumptions. Ranked-list methods are popular non-parametric models and capture individuals’ behavior by associating them with preference lists of options sorted in decreasing order of preference. Individuals are assumed always to choose the option best placed in their preference list when confronted with an alternative. Despite the generality and simplicity of ranked-list methods, a major drawback is the exponential increase in the number of potential lists. Column generation (CG) can be employed to address this issue, with the CG subproblem modeled as a generalized linear ordering problem (GLOP). In this talk, we present a dynamic programming algorithm to solve GLOPs. The proposed method is generic and capable of handling different settings without requiring drastic changes to its implementation. The proposed algorithm outperforms a previously proposed Branch-and-Cut algorithm. Our algorithm efficiently generates preference lists when incorporated into maximum likelihood and minimum L1 estimators. The algorithm performs well when facing instances with numerous observations, which is crucial as non-parametric choice models heavily rely on data volume for accurate estimations. This is joint work with Claudio Contardo (Concordia University), Gerardo Berbeglia (Melbourne Business School, The University of Melbourne), and Jean-François Cordeau (HEC Montréal).
Zoom link: https://unipd.zoom.us/j/81999343404
Meeting ID: 819 9934 3404
I'm sorry for cross-posting.
On Monday, May 26, we will have the pleasure of having Luciano Costa as a guest at the Department of Information Engineering (DEI) of the University of Padova. He is an Assistant Professor at the Universidade Federal de Paraíba, Brazil.
On Monday, May 26, we will have the pleasure of having Luciano Costa as a guest at the Department of Information Engineering (DEI) of the University of Padova. He is an Assistant Professor at the Universidade Federal de Paraíba, Brazil.
Luciano will hold a seminar at DEI, in the meeting room on the 4th floor. Nonetheless, the seminar will be held in a hybrid mode on Zoom to make it accessible to other interested people. You can find the title and abstract below.
We hope you'll be able to join us either in person or remotely.
Best,
Alice
---------------------
Title: Column generation applied to the estimation of non-parametric discrete-choice models
Abstract: Discrete choice models (DCMs) provide probabilities for individuals choosing a certain alternative when faced with a set of limited options. DCMs can be parametric or non-parametric. Parametric models are easier to estimate but require assumptions about individuals' preferences, while non-parametric models rely solely on training data without any assumptions. Ranked-list methods are popular non-parametric models and capture individuals’ behavior by associating them with preference lists of options sorted in decreasing order of preference. Individuals are assumed always to choose the option best placed in their preference list when confronted with an alternative. Despite the generality and simplicity of ranked-list methods, a major drawback is the exponential increase in the number of potential lists. Column generation (CG) can be employed to address this issue, with the CG subproblem modeled as a generalized linear ordering problem (GLOP). In this talk, we present a dynamic programming algorithm to solve GLOPs. The proposed method is generic and capable of handling different settings without requiring drastic changes to its implementation. The proposed algorithm outperforms a previously proposed Branch-and-Cut algorithm. Our algorithm efficiently generates preference lists when incorporated into maximum likelihood and minimum L1 estimators. The algorithm performs well when facing instances with numerous observations, which is crucial as non-parametric choice models heavily rely on data volume for accurate estimations. This is joint work with Claudio Contardo (Concordia University), Gerardo Berbeglia (Melbourne Business School, The University of Melbourne), and Jean-François Cordeau (HEC Montréal).
Zoom link: https://unipd.zoom.us/j/81999343404
Meeting ID: 819 9934 3404
Reply all
Reply to author
Forward
0 new messages