000262662 001__ 262662
000262662 005__ 20190619220145.0
000262662 02470 $$a1812.09747$$2ArXiv
000262662 037__ $$aARTICLE
000262662 245__ $$aLet Me Not Lie: Learning MultiNomial Logit
000262662 260__ $$c2018-12-23
000262662 269__ $$a2018-12-23
000262662 336__ $$aJournal Articles
000262662 520__ $$aDiscrete choice models generally assume that model specification is known a priori. In practice, determining the utility specification for a particular application remains a difficult task and model misspecification may lead to biased parameter estimates. In this paper, we propose a new mathematical framework for estimating choice models in which the systematic part of the utility specification is divided into an interpretable part and a learning representation part that aims at automatically discovering a good utility specification from available data. We show the effectiveness of our framework by augmenting the utility specification of the Multinomial Logit Model (MNL) with a new non-linear representation arising from a Neural Network (NN). This leads to a new choice model referred to as the Learning Multinomial Logit (L-MNL) model. Our experiments show that our L-MNL model outperformed the traditional MNL models and existing hybrid neural network models both in terms of predictive performance and accuracy in parameter estimation.
000262662 6531_ $$adiscrete choice model
000262662 6531_ $$arepresentation learning,
000262662 6531_ $$aneural network
000262662 6531_ $$adeep learning
000262662 700__ $$aSifringer, Brian
000262662 700__ $$g276768$$aLurkin, Virginie$$0250377
000262662 700__ $$0242925$$aAlahi, Alexandre$$g129343
000262662 773__ $$tarxiv
000262662 8560_ $$falain.borel@epfl.ch
000262662 8564_ $$uhttps://infoscience.epfl.ch/record/262662/files/1812.09747.pdf$$s3160534
000262662 909C0 $$xU13529$$pVITA$$malexandre.alahi@epfl.ch$$zPasquier, Simon$$0252606
000262662 909CO $$qGLOBAL_SET$$particle$$pENAC$$ooai:infoscience.epfl.ch:262662
000262662 960__ $$aalexandre.alahi@epfl.ch
000262662 961__ $$afantin.reichler@epfl.ch
000262662 973__ $$aEPFL$$sSUBMITTED$$rREVIEWED
000262662 980__ $$aARTICLE
000262662 981__ $$aoverwrite