pith. sign in

arxiv: 2606.13418 · v1 · pith:ATKG26NOnew · submitted 2026-06-11 · 🧮 math.OC

Coordinate-wise Polyhedral Method for Eliciting Multivariate Linear Utility and Univariate Nonlinear Utility Functions

Jiaxin Wei , Jia Liu , Huifu Xu This is my paper

Pith reviewed 2026-06-27 05:55 UTC · model grok-4.3

classification 🧮 math.OC
keywords polyhedral methodutility elicitationlinear convergencecoordinate-wise cutsKantorovich distancepiecewise linear approximationpairwise comparisons
0
0 comments X

The pith

A coordinate-wise polyhedral method reduces the uncertainty polyhedron for utility functions at a linear convergence rate.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes a coordinate-wise polyhedral method (CPM) for eliciting utility functions from pairwise comparisons. Instead of solving coupled optimization problems for queries, it pre-specifies coordinate-wise cuts and solves a linear system to design the queries. This approach guarantees that the diameter of the polyhedron shrinks at a linear rate. The method extends to nonlinear univariate utilities by approximating them with piecewise linear functions and using an adaptive breakpoint strategy, showing that the Kantorovich distance to the true function also converges linearly.

Core claim

Under the CPM framework, the diameter of the polyhedron is reduced at a linear convergence rate. For univariate piecewise-linear utility functions represented by their increment vectors, the Kantorovich distance between the true utility and the estimated one decreases at a linear convergence rate. The method further extends to general nondecreasing Lipschitz continuous utilities via piecewise-linear approximation with an adaptive-breakpoint strategy that ensures convergence of the ambiguity set to the true function as the number of queries increases.

What carries the argument

The coordinate-wise polyhedral method (CPM), which pre-specifies coordinate-wise cuts in advance and designs pairwise comparison queries by solving a linear system of equations.

If this is right

  • The diameter of the polyhedron for multivariate linear utility functions decreases linearly with the number of queries.
  • The Kantorovich distance for univariate nonlinear utility functions decreases at a linear rate.
  • The set of piecewise-linear utility functions converges to the true nondecreasing Lipschitz utility as queries increase.
  • An explicit bound for the approximation error is derived.
  • Numerical experiments confirm stable convergence comparable to standard methods.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • CPM may reduce computational cost compared to methods that solve optimization problems for each query since it uses linear systems.
  • The adaptive breakpoint strategy could be tested in other approximation settings for utility functions.
  • Linear convergence might enable efficient real-time elicitation in decision support systems.

Load-bearing premise

Pairwise comparison queries can always be obtained without inconsistencies by solving a linear system after pre-specifying the coordinate-wise cuts.

What would settle it

A case where solving the linear system for query design leads to inconsistent queries or where the polyhedron diameter fails to decrease linearly with successive cuts would disprove the convergence claim.

Figures

Figures reproduced from arXiv: 2606.13418 by Huifu Xu, Jia Liu, Jiaxin Wei.

Figure 1
Figure 1. Figure 1: Example of a query for funds The main challenge in the procedure is in the design of queries that enable the modeler to reduce the ambiguity set efficiently. Toubia et al. [2004, 2007] select queries by exploiting the geometry of the current polyhedron through its analytic center. They propose an optimal query design framework that is guided by four practical criteria including nondominance, feasibility, c… view at source ↗
Figure 2
Figure 2. Figure 2: The process of utility elicitation between the modeler and the DM [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Procedures of CPM. (a) Find the MEB center, the longest axis of MOE and orthogonal directions. (b) Find query and implement cutting. (c) Find coordinate-wise cutting plane and the query. separating hyperplane of the query (A, B), where x A N − x B N = − N X−1 i=1 αici , (20a) x A i − x B i = (x A N − x B N ) + αi = − N X−1 i=1 αici + αi , for i = 1, . . . , N − 1. (20b) Remark 3.1 (Criterion of determining… view at source ↗
Figure 4
Figure 4. Figure 4: MEB vs Sonnevend’s ellipsoid a better estimate in the case where the true partworth vector is located at the boundary of P k,i, see [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Illustration of MEBs and cutting planes in major iteration one. (a) MEBs after each coordinate-wise cut in R 2 . (b) Cutting plane in R 3 after one coordinate-wise cut. (c) Cutting plane in R 3 after second coordinate-wise cut. (a) (b) (c) [PITH_FULL_IMAGE:figures/full_fig_p017_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Illustration of MEBs and cutting planes in major iteration two. (a) MEBs after each coordinate cut in R 2 . (b) Cutting plane in R 3 after one coordinate-wise cut. (c) Cutting plane in R 3 after second coordinate-wise cut. 4 Univariate piecewise-linear utility function In the domain of individual decision-making under risk, preferences are often modeled using nonlinear univariate utility or distortion func… view at source ↗
Figure 7
Figure 7. Figure 7: The cutting process Yaari’s dual theory [Yaari, 1987], rank-dependent expected utility (RDEU) [Quiggin, 1982, Wakker, 1994], and cumulative prospect theory (CPT) [Tversky and Kahneman, 1992] allow for nonlinear transformations of both outcomes and probabilities. These models often incorporate univariate nonlinear utility and/or distortion functions to represent individual attitudes toward risk, loss, and p… view at source ↗
Figure 8
Figure 8. Figure 8: (a) The range of g(·). (b) The range of GA and −GB. (c) The range of GA − GB. Next, we present the convergence of Algorithm 2. Unlike the multivariate linear utility function, we need to consider the convergence of the utility functions estimated from the polyhedron elicited from the increment vectors. Theorem 4.2 (Convergence of Algorithm 2 in terms of utility functions). Let X = {x1 < · · · < xN+1} ⊆ [x,… view at source ↗
Figure 9
Figure 9. Figure 9: (a) depicts changes in the ranges of the utility function values hinf u∈Uk,i X u(x),supu∈Uk,i X u(x) i over its domain [0, 4] as the number of queries (N.q) increases. We can see that as the range of the utility functions shrinks to the true utility value pointwise, so does the utility function induced by the MEB center [PITH_FULL_IMAGE:figures/full_fig_p026_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: The cutting process 5 Extension to general continuous utility functions In Section 4, the true utility is assumed to be piecewise-linear with a fixed set of breakpoints. In this section we extend the framework to the case where the DM’s true utility is a general continuous utility function on a compact interval. Assumption 4. The decision maker’s preference is represented by a utility function u ∗ : [x, x… view at source ↗
Figure 11
Figure 11. Figure 11: The range of u(ˆxi) in polyhedra. Upper row: changes of the ranges of U k,i X at u(ˆx1), u(ˆx2) and u(ˆx3) as the number of queries increases. Lower row: Evolution of the upper bound for u(ˆx1), u(ˆx2) and u(ˆx3). In the discussions below, we abuse terminology slightly by referring to X k interchangeably as the grid and hk mesh size. This is sensible from the numerical approximation perspective. The updat… view at source ↗
Figure 12
Figure 12. Figure 12: Test 1 with X 0 (1) = {0, 0.6, 2, 4}. (a) Worst-case utility function arg maxu∈Uk dlK(u ∗ , u). (b) Range of utility function values corresponding to U k . (a) (b) [PITH_FULL_IMAGE:figures/full_fig_p038_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Test 2 with X 0 (2) = {0, 3, 3.5, 4}. (a) Worst-case utility function arg maxu∈Uk dlK(u ∗ , u). (b) Range of utility function values corresponding to U k . over a set of metrics introduced below. We consider customers’ preferences on a wireless headset product. The product is characterized by four attributes: battery life (hours per charge) xbattery, wearing comfort xcomfort, active noise cancellation (AN… view at source ↗
Figure 14
Figure 14. Figure 14: Comparative analysis of CPM and POLY over 100 customers (solid curves indicate the means and shaded areas indicate the ranges). (a) Partworth sign correctness based on the AC/MEB centers. (b) Worst-case partworth sign correctness. (c) Utility regret based on the AC/MEB centers. (d) Worst-case utility regret. (e) Polyhedron diameter. 41 [PITH_FULL_IMAGE:figures/full_fig_p041_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Average price gap across 100 DMs after each query [PITH_FULL_IMAGE:figures/full_fig_p043_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Analytic centers and MEB centers of some examples of polygons [PITH_FULL_IMAGE:figures/full_fig_p057_16.png] view at source ↗
read the original abstract

In this paper, we propose a coordinate-wise polyhedral method (CPM) for cutting polyhedra with theoretical guarantees of convergence. Unlike the existing polyhedral method, which designs pairwise comparison queries by solving coupled optimization problems and performs cuts subsequently, CPM specifies coordinate-wise cuts in advance and then designs corresponding pairwise comparison queries by solving a linear system of equations. Under this framework, we show that CPM reduces the diameter of the polyhedron at a linear convergence rate. Moreover, we extend CPM to a univariate piecewise-linear utility function by representing it with its increment vector over consecutive breakpoints. We show that the Kantorovich distance between the true utility function and the estimated one obtained by CPM decreases at a linear convergence rate. Further, we extend the CPM to general nondecreasing Lipschitz continuous utility functions by piecewise-linear approximation (PLA). We introduce an adaptive-breakpoint strategy to avoid direction errors caused by PLA. We prove that the set of piecewise-linear utility functions corresponding to the ambiguity set of increment vectors converges to the true utility function as the number of queries increases, and derive an explicit bound for the approximation. Finally, to evaluate the performance of CPM, we conduct a series of numerical experiments. The results demonstrate comparable convergence to the standard polyhedral method in the case of linear multivariate utility functions. For nonlinear univariate utility functions, CPM achieves stable convergence to the true utility function, in line with the theoretical findings.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes a coordinate-wise polyhedral method (CPM) for eliciting multivariate linear utility functions by pre-specifying coordinate-wise cuts and designing pairwise comparison queries via a linear system of equations. It claims this yields linear convergence in polyhedron diameter reduction. The method is extended to univariate piecewise-linear utilities (with linear convergence in Kantorovich distance between true and estimated functions) and to general nondecreasing Lipschitz utilities via piecewise-linear approximation (PLA) with an adaptive breakpoint strategy that ensures convergence of the ambiguity set with an explicit bound. Numerical experiments compare CPM to the standard polyhedral method.

Significance. If the linear-system solvability and convergence arguments hold, CPM offers a structured alternative to coupled optimization-based polyhedral methods with explicit linear rates, which would be useful for preference elicitation in decision analysis and optimization. The PLA extension with adaptive breakpoints and Kantorovich-distance bound adds value for nonlinear cases.

major comments (2)
  1. [CPM Framework (prior to the diameter-reduction theorem)] The linear diameter reduction (abstract) and all subsequent rates rest on the claim that, after pre-specifying coordinate-wise cuts, the linear system for pairwise-query design always admits a solution that produces a valid cut reducing diameter by a fixed factor <1. No explicit proof, rank condition, or consistency argument for this system across iterations appears in the framework description; if the system is inconsistent or the query fails to implement the intended cut due to cross-coordinate coupling, the linear-rate claim fails.
  2. [Univariate Nonlinear and PLA Extensions] The Kantorovich-distance linear convergence for the univariate piecewise-linear case and the PLA convergence proof for Lipschitz utilities inherit the same linear-system assumption; any gap in solvability or diameter reduction propagates directly to these extensions and to the adaptive-breakpoint strategy.
minor comments (2)
  1. [Numerical Experiments] The numerical experiments claim 'comparable convergence' and 'stable convergence'; add quantitative tables or plots with explicit iteration counts, diameter values, and distance metrics to allow direct comparison with the theoretical rates.
  2. [CPM Framework] Clarify the precise form of the linear system (variables, constraints, and right-hand side) used to design the queries once cuts are pre-specified.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thorough review and for identifying the need for explicit justification of the linear-system solvability. We agree that a dedicated proof of consistency, rank conditions, and diameter reduction is required to support the linear-rate claims and their extensions. We will revise the manuscript accordingly by adding this material prior to the main theorems. Point-by-point responses follow.

read point-by-point responses

  1. Referee: [CPM Framework (prior to the diameter-reduction theorem)] The linear diameter reduction (abstract) and all subsequent rates rest on the claim that, after pre-specifying coordinate-wise cuts, the linear system for pairwise-query design always admits a solution that produces a valid cut reducing diameter by a fixed factor <1. No explicit proof, rank condition, or consistency argument for this system across iterations appears in the framework description; if the system is inconsistent or the query fails to implement the intended cut due to cross-coordinate coupling, the linear-rate claim fails.

    Authors: We appreciate this observation. The current framework description states that the linear system is solved to design queries but does not contain a self-contained argument establishing solvability, full rank of the system matrix, or invariance of the reduction factor across iterations. This omission leaves open the possibility of inconsistency due to coordinate coupling. We will insert a new subsection immediately before the diameter-reduction theorem that proves: (i) the coordinate-wise cut specification yields a square, full-rank matrix independent of prior cuts; (ii) the unique solution corresponds to a valid separating hyperplane; and (iii) the resulting cut reduces the polyhedron diameter by a fixed factor strictly less than one. This addition directly addresses the referee’s concern. revision: yes

  2. Referee: [Univariate Nonlinear and PLA Extensions] The Kantorovich-distance linear convergence for the univariate piecewise-linear case and the PLA convergence proof for Lipschitz utilities inherit the same linear-system assumption; any gap in solvability or diameter reduction propagates directly to these extensions and to the adaptive-breakpoint strategy.

    Authors: The referee is correct that the convergence statements for the piecewise-linear univariate case (Kantorovich distance) and the PLA extension (ambiguity-set convergence with explicit bound) rest on the base CPM linear-system properties. Once the new subsection establishes solvability and the fixed-factor diameter reduction for the linear multivariate setting, the same arguments carry over verbatim to the increment-vector representation used for univariate utilities and to the adaptive-breakpoint construction. We will add a short remark in each extension section that explicitly invokes the new proof, thereby closing the inheritance gap without altering the adaptive strategy itself. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained with independent convergence proofs.

full rationale

The paper defines CPM by pre-specifying coordinate-wise cuts then solving a linear system for queries, and separately proves linear diameter reduction and Kantorovich convergence under that framework. No quoted step reduces a claimed prediction or theorem to a fitted input, self-citation, or definitional tautology. The extensions to univariate nonlinear and PLA cases inherit the same framework but add explicit adaptive strategies and approximation bounds that are derived rather than presupposed. This matches the default expectation of a non-circular paper.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Based solely on the abstract, the claims rest on the domain assumption that utility functions are nondecreasing and Lipschitz continuous (for the general case) and that the linear system for query design is always solvable; no free parameters or invented entities are mentioned.

axioms (1)
  • domain assumption Utility function is nondecreasing and Lipschitz continuous for the general nondecreasing case extension via PLA.
    Invoked in the abstract when extending CPM to general utility functions and proving convergence of the ambiguity set.

pith-pipeline@v0.9.1-grok · 5786 in / 1264 out tokens · 22830 ms · 2026-06-27T05:55:08.660236+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

73 extracted references · 3 canonical work pages · 1 internal anchor

  1. [1]

    Journal of the Royal Statistical Society: Series B , volume =

    Optimum Experimental Designs , author =. Journal of the Royal Statistical Society: Series B , volume =

  2. [2]

    2006 , address =

    Optimal Design of Experiments , author =. 2006 , address =

    2006

  3. [3]

    Mathematical Programming , pages=

    Eliciting Von Neumann--Morgenstern utility from discrete choices with response error , author=. Mathematical Programming , pages=. 2026 , publisher=

    2026

  4. [4]

    Journal of Marketing Research , volume=

    The effectiveness of alternative preference elicitation procedures in predicting choice , author=. Journal of Marketing Research , volume=. 1993 , publisher=

    1993

  5. [5]

    Operations Research , volume=

    Utility preference robust optimization with moment-type information structure , author=. Operations Research , volume=. 2024 , publisher=

    2024

  6. [6]

    Omega , volume=

    Filling in pattern designs for incomplete pairwise comparison matrices:(quasi-) regular graphs with minimal diameter , author=. Omega , volume=. 2022 , publisher=

    2022

  7. [7]

    Proceedings of the 12th ACM Conference on Recommender Systems , pages=

    Preference elicitation as an optimization problem , author=. Proceedings of the 12th ACM Conference on Recommender Systems , pages=

  8. [8]

    Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining , pages=

    Towards conversational recommender systems , author=. Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining , pages=

  9. [9]

    Marketing Science , volume=

    Construction of heterogeneous conjoint choice designs: A new approach , author=. Marketing Science , volume=. 2015 , publisher=

    2015

  10. [10]

    Journal of Marketing Research , volume=

    Heterogeneous conjoint choice designs , author=. Journal of Marketing Research , volume=. 2005 , publisher=

    2005

  11. [11]

    Journal of Business & Economic Statistics , volume=

    An efficient algorithm for constructing Bayesian optimal choice designs , author=. Journal of Business & Economic Statistics , volume=. 2009 , publisher=

    2009

  12. [12]

    Journal of Marketing Research , volume=

    A comparison of criteria to design efficient choice experiments , author=. Journal of Marketing Research , volume=. 2006 , publisher=

    2006

  13. [13]

    Marketing Science , volume=

    Profile construction in experimental choice designs for mixed logit models , author=. Marketing Science , volume=. 2002 , publisher=

    2002

  14. [14]

    Journal of Marketing Research , volume=

    Designing conjoint choice experiments using managers' prior beliefs , author=. Journal of Marketing Research , volume=. 2001 , publisher=

    2001

  15. [15]

    Journal of Marketing Research , volume=

    Optimal design for multinomial choice experiments , author=. Journal of Marketing Research , volume=. 2002 , publisher=

    2002

  16. [16]

    SIAM Journal on Optimization , volume=

    Two algorithms for the minimum enclosing ball problem , author=. SIAM Journal on Optimization , volume=. 2008 , publisher=

    2008

  17. [17]

    Applied Sciences , volume=

    AI-driven recommendations: A systematic review of the state of the art in e-commerce , author=. Applied Sciences , volume=. 2023 , publisher=

    2023

  18. [18]

    The Palgrave handbook of interactive marketing , pages=

    AI-based recommendation systems: the ultimate solution for market prediction and targeting , author=. The Palgrave handbook of interactive marketing , pages=. 2023 , publisher=

    2023

  19. [19]

    Advanced Engineering Informatics , volume=

    An AI-based open recommender system for personalized labor market driven education , author=. Advanced Engineering Informatics , volume=. 2022 , publisher=

    2022

  20. [20]

    Mathematical Programming , pages=

    Preference ambiguity and robustness in multistage decision making , author=. Mathematical Programming , pages=. 2025 , publisher=

    2025

  21. [21]

    Behavioral Science , volume=

    Measuring utility by a single-response sequential method , author=. Behavioral Science , volume=. 1964 , publisher=

    1964

  22. [22]

    Management Science , volume=

    Loss aversion under prospect theory: A parameter-free measurement , author=. Management Science , volume=. 2007 , publisher=

    2007

  23. [23]

    Management Science , volume=

    Parameter-free elicitation of utility and probability weighting functions , author=. Management Science , volume=. 2000 , publisher=

    2000

  24. [24]

    Management Science , volume=

    Eliciting von Neumann-Morgenstern utilities when probabilities are distorted or unknown , author=. Management Science , volume=. 1996 , publisher=

    1996

  25. [25]

    Econometrica: Journal of the Econometric Society , pages=

    Investigating generalizations of expected utility theory using experimental data , author=. Econometrica: Journal of the Econometric Society , pages=. 1994 , publisher=

    1994

  26. [26]

    Journal of Risk and Uncertainty , volume=

    Nonlinear weighting of probabilities and violations of the betweenness axiom , author=. Journal of Risk and Uncertainty , volume=

  27. [27]

    Journal of Risk and Uncertainty , volume=

    Advances in prospect theory: Cumulative representation of uncertainty , author=. Journal of Risk and Uncertainty , volume=. 1992 , publisher=

    1992

  28. [28]

    , author=

    Weighing risk and uncertainty. , author=. Psychological Review , volume=. 1995 , publisher=

    1995

  29. [29]

    Theory and Decision , volume=

    Separating marginal utility and probabilistic risk aversion , author=. Theory and Decision , volume=. 1994 , publisher=

    1994

  30. [30]

    Journal of Economic Behavior & Organization , volume=

    A theory of anticipated utility , author=. Journal of Economic Behavior & Organization , volume=. 1982 , publisher=

    1982

  31. [31]

    Econometrica: Journal of the Econometric Society , pages=

    The dual theory of choice under risk , author=. Econometrica: Journal of the Econometric Society , pages=. 1987 , publisher=

    1987

  32. [32]

    1947 , publisher=

    Theory of games and economic behavior, 2nd rev , author=. 1947 , publisher=

    1947

  33. [33]

    Mathematics of Operations Research , year=

    On the convex formulations of robust Markov decision processes , author=. Mathematics of Operations Research , year=

  34. [34]

    Mathematics of Operations Research , volume=

    Robust dynamic programming , author=. Mathematics of Operations Research , volume=. 2005 , publisher=

    2005

  35. [35]

    1994 , publisher=

    Markov decision processes: discrete stochastic dynamic programming , author=. 1994 , publisher=

    1994

  36. [36]

    Science , volume=

    Dynamic programming , author=. Science , volume=. 1966 , publisher=

    1966

  37. [37]

    Mathematical Programming , pages=

    Distributional utility preference robust optimization models in multi-attribute decision making , author=. Mathematical Programming , pages=. 2024 , publisher=

    2024

  38. [38]

    Adaptive preference elicitation in preference robust CPT-based shortfall , author=. , year=

  39. [39]

    arXiv preprint arXiv:2503.23269 , year=

    Modified Polyhedral Method for Elicitation of Shape-Free Utility and Conservatism Reduction in Robust Optimization , author=. arXiv preprint arXiv:2503.23269 , year=

    work page arXiv

  40. [40]

    Management Science , volume=

    Dynamic experiments for estimating preferences: An adaptive method of eliciting time and risk parameters , author=. Management Science , volume=. 2013 , publisher=

    2013

  41. [41]

    Marketing Science , volume=

    Fast polyhedral adaptive conjoint estimation , author=. Marketing Science , volume=. 2003 , publisher=

    2003

  42. [42]

    Journal of Marketing Research , volume=

    Polyhedral methods for adaptive choice-based conjoint analysis , author=. Journal of Marketing Research , volume=. 2004 , publisher=

    2004

  43. [43]

    Transportation Research Part B: Methodological , volume=

    A comparison of different Bayesian design criteria for setting up stated preference studies , author=. Transportation Research Part B: Methodological , volume=. 2012 , publisher=

    2012

  44. [44]

    Doklady Akademii Nauk , volume=

    A polynomial algorithm in linear programming , author=. Doklady Akademii Nauk , volume=. 1979 , organization=

    1979

  45. [45]

    Lecture Notes from Stanford University , year=

    Analytic center cutting-plane method , author=. Lecture Notes from Stanford University , year=

  46. [46]

    Marketing Science , volume=

    Probabilistic polyhedral methods for adaptive choice-based conjoint analysis: Theory and application , author=. Marketing Science , volume=. 2007 , publisher=

    2007

  47. [47]

    Artificial Intelligence , volume=

    Label ranking by learning pairwise preferences , author=. Artificial Intelligence , volume=. 2008 , publisher=

    2008

  48. [48]

    International Symposium on Intelligent Data Analysis , pages=

    A policy iteration algorithm for learning from preference-based feedback , author=. International Symposium on Intelligent Data Analysis , pages=. 2013 , organization=

    2013

  49. [49]

    , author=

    Preference-learning based Inverse Reinforcement Learning for Dialog Control. , author=. INTERSPEECH , volume=

  50. [50]

    2000 , publisher=

    Stated choice methods: analysis and applications , author=. 2000 , publisher=

    2000

  51. [51]

    Conditional logit analysis of qualitative choice behavior , author=

  52. [52]

    American Economic Review , volume=

    Economic choices , author=. American Economic Review , volume=. 2001 , publisher=

    2001

  53. [53]

    Biometrika , volume=

    A continuum of paired comparisons models , author=. Biometrika , volume=. 1990 , publisher=

    1990

  54. [54]

    Journal of Eeconometrics , volume=

    Marketing models of consumer heterogeneity , author=. Journal of Eeconometrics , volume=. 1998 , publisher=

    1998

  55. [55]

    Journal of Consumer Research , volume=

    Improving parameter estimates and model prediction by aggregate customization in choice experiments , author=. Journal of Consumer Research , volume=. 2001 , publisher=

    2001

  56. [56]

    Operations Research , volume=

    Learning preferences under noise and loss aversion: An optimization approach , author=. Operations Research , volume=. 2013 , publisher=

    2013

  57. [57]

    Operations Research , volume=

    Ellipsoidal methods for adaptive choice-based conjoint analysis , author=. Operations Research , volume=. 2019 , publisher=

    2019

  58. [58]

    Journal of Marketing Research , volume=

    Efficient experimental design with marketing research applications , author=. Journal of Marketing Research , volume=. 1994 , publisher=

    1994

  59. [59]

    Journal of Marketing research , volume=

    The importance of utility balance in efficient choice designs , author=. Journal of Marketing research , volume=. 1996 , publisher=

    1996

  60. [60]

    Management Science , volume=

    Decision making under uncertainty when preference information is incomplete , author=. Management Science , volume=. 2015 , publisher=

    2015

  61. [61]

    Financial Innovation , volume=

    Personalized fund recommendation with dynamic utility learning , author=. Financial Innovation , volume=. 2025 , publisher=

    2025

  62. [62]

    Advances in Neural Information Processing Systems , volume=

    Active preference learning with discrete choice data , author=. Advances in Neural Information Processing Systems , volume=

  63. [63]

    A Cost-Effective Framework for Preference Elicitation and Aggregation

    A cost-effective framework for preference elicitation and aggregation , author=. arXiv preprint arXiv:1805.05287 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  64. [64]

    Ijcai , volume=

    Incremental utility elicitation with the minimax regret decision criterion , author=. Ijcai , volume=

  65. [65]

    Artificial Intelligence , volume=

    Constraint-based optimization and utility elicitation using the minimax decision criterion , author=. Artificial Intelligence , volume=. 2006 , publisher=

    2006

  66. [66]

    Advances in Neural Information Processing Systems , volume=

    A bayesian approach for policy learning from trajectory preference queries , author=. Advances in Neural Information Processing Systems , volume=

  67. [67]

    ICRA Workshop on autonomous learning , volume=

    Preference-based evolutionary direct policy search , author=. ICRA Workshop on autonomous learning , volume=

  68. [68]

    Machine learning , volume=

    Preference-based reinforcement learning: evolutionary direct policy search using a preference-based racing algorithm , author=. Machine learning , volume=. 2014 , publisher=

    2014

  69. [69]

    arXiv preprint arXiv:2003.01899 , year=

    Robust active preference elicitation , author=. arXiv preprint arXiv:2003.01899 , year=

    work page arXiv 2003

  70. [70]

    Operations Research , volume=

    A constructive approach to estimating pure characteristics demand models with pricing , author=. Operations Research , volume=. 2015 , publisher=

    2015

  71. [71]

    1993 , publisher=

    Automobile prices in market equilibrium: Part I and II , author=. 1993 , publisher=

    1993

  72. [72]

    Optimisation Methods and Software , volume=

    Linear convergence of a modified Frank--Wolfe algorithm for computing minimum-volume enclosing ellipsoids , author=. Optimisation Methods and Software , volume=. 2008 , publisher=

    2008

  73. [73]

    2016 , publisher=

    Minimum-volume ellipsoids: Theory and algorithms , author=. 2016 , publisher=

    2016