New algorithms for discrete choice models using semi-supervised learning.
problem Calibrating discrete choice models with limited labeled data.
method Adapted and developed semi-supervised learning algorithms for choice modeling.
result New algorithms improve prediction accuracy and computational efficiency.
Active learning recovers choice model from noisy data.
problem Identifying non-parametric choice models from noisy data.
method Directed acyclic graph (DAG) representation and inclusion-exclusion approach.
result Algorithm more accurately recovers frequent preferences.
Revisits PPO design choices, exposing failure modes and proposing alternatives.
problem Failure modes of standard PPO in new environments.
method Revisits standard PPO design choices, exposes failure modes, and proposes alternative approaches.
result Alternative design choices prevent failure modes in new environments.
Optimizes choice sets to influence group decisions.
problem Maximizing agreement or disagreement in group decisions.
method Discrete choice modeling to develop optimization framework.
result Promoting a choice can be easier than encouraging consensus or discord.
Estimates multi-attribute choice preferences using private signals and matrix factorization.
problem Modeling multi-attribute choice preferences under weak assumptions.
method Generative choice model with latent factor matrices and private signals; multi-stage matrix factorization.
result Validated estimation performance of novel algorithm through simulations.
Proposes new methods for Markov chain choice models with panel data.
problem Dependence among transactions for the same customer in historical data.
method Expectation-maximization (EM) algorithms incorporating partial-ordering preference information.
result EM algorithms outperform traditional methods on synthetic and real datasets.
Study binary choice with asymmetric loss, offering simple solutions.
problem Binary choice with asymmetric loss in data-rich environments.
method Loss-based reweighting of logistic regression or machine learning techniques.
result Valid decisions on binary outcomes with general loss functions.
Bayesian active learning finds individual's most preferred choice with deep Gaussian processes.
problem Finding individual's most preferred choice through pairwise comparisons.
method Active learning scheme using probabilistic models based on choice models and deep Gaussian processes, with a novel acquisition function.
result Effectiveness of the proposed active learning algorithm and models as demonstrated by experiments.
The paper connects discrete choice models to multi-armed bandit algorithms with sublinear regret bounds.
problem Optimizing user choices in a multi-armed bandit setting.
method Establishes connections between discrete choice models and multi-armed bandit algorithms, providing sublinear regret bounds and novel algorithms.
result Sublinear regret bounds for a family of algorithms, including the Exp3 algorithm.
Simple algorithms identify best items or full rankings from choice-based feedback.
problem Learning to identify the best item or full ranking from choice-based feedback.
method Nested Elimination (NE) and Nested Partition (NP) algorithms.
result NE is worst-case asymptotically optimal, NP is optimal up to a constant factor.
Study improves choice model accuracy and heterogeneity representation using mixture models.
problem Improving prediction accuracy and heterogeneity representation in choice models.
method Semi-nonparametric Latent Class Choice Model with mixture models and EM algorithm.
result Mixture models enhance prediction accuracy and heterogeneity representation without sacrificing interpretability.
The paper proposes a new method to learn choice functions using Pareto-embeddings.
problem Learning subset choices from feature vectors.
method Embedding choice alternatives into a higher-dimensional utility space and identifying choice sets with Pareto-optimal points. Minimizing a differentiable loss function.
result The feasibility of learning a Pareto-embedding demonstrated on benchmark datasets.
Optimizes Metropolis-Hastings algorithms for efficient sampling in high dimensions.
problem Efficiently sampling from complex target distributions in high-dimensional spaces.
method Analyzes and optimizes the Barker proposal and other locally-balanced algorithms.
result Derives optimal noise distribution and balancing function for the Barker proposal.
Robo-advisor learns investor's risk preference through portfolio choices.
problem Learning investors' risk preferences without prior knowledge.
method Reinforcement learning framework with exploration-exploitation algorithm.
result Algorithm's value function converges to optimal over polynomial periods.
Paper extends top-k Mallows model for better user preference analysis.
problem Capturing real-world user preferences focusing on a limited set of items.
method Generalized top-k Mallows model, novel sampling scheme, efficient algorithm, active learning.
result New tools for analysis and prediction in decision-making scenarios.
Examines how algorithms affect user autonomy and information choice.
problem Impact of algorithmic recommendations on user autonomy and free choice.
method Double dichotomy analysis of user intentions and actions, prior and posterior information rearrangement.
result Algorithms can expand or limit user cognitive and social horizons.
A new method reduces complexity in estimating dynamic choice models.
problem Estimating structural parameters in dynamic discrete choice models using behavioral data.
method Two-stage approach: inverse reinforcement learning for Q-function estimation, state selection via clustering, and maximum likelihood estimation with nested fixed-point algorithm.
result The method mitigates the curse of dimensionality and provides finite-sample bounds on estimation error.
Paper recovers top-two answers and confusion probability in multi-choice crowdsourcing.
problem Recovering top-two answers and confusion probability in multi-choice crowdsourcing tasks.
method Proposes a two-stage inference algorithm based on a model quantifying task difficulty and worker reliability.
result Achieves minimax optimal convergence rate and outperforms other algorithms in synthetic and real data experiments.
Binary choice forests model customer choices in retailing.
problem Estimating DCMs using transaction data is challenging and prone to misspecification.
method Random forest of binary decision trees to represent DCMs, interpretable and consistent predictions.
result Random forest can predict choice probabilities and assortments unseen in training data.
Study investigates key design choices in on-policy RL algorithms.
problem Lack of transparency in RL algorithm implementations.
method Implemented >50 design choices in a unified RL framework.
result Insights and recommendations for on-policy RL training.
Graph neural networks improve residential location choice predictions.
problem Capturing spatial dependence in discrete choice models.
method Graph Neural Networks (GNN) for analyzing spatial alternatives.
result GNN-DCMs outperform classical models in residential location choice predictions.
A general class of Newton algorithms on Graßmann and Lagrange-Graßmann manifolds is introduced, that depends on an arbitrary pair of local coordinates. Local quadratic convergence of the algorithm is shown under a suitable condition on the choice of coordinate systems. Our result extends and unifies previous convergenc…
Study identifies Markov chain model parameters from small assortments.
problem Identifying parameters of Markov chain choice models from large assortments.
method Simple and efficient algorithm to recover parameters from assortments of sizes two and three.
result Parameters of the Markov chain choice model can be identified from assortments of sizes two and three.
Two algorithms optimize assortment selection for user choices in unknown MNL models.
problem Sequential assortment selection with unknown multinomial logit parameters.
method Upper confidence bound algorithms for MNL contextual bandits.
result Optimal regret bounds for assortment selection problems.
Random Machines improves SVM performance with free kernel choice.
problem Efficiency and accuracy in solving classification and regression problems.
method Bagged-weighted support vector model with free kernel choice.
result Improved accuracy and reduced computational time.
SpectralTS improves efficiency of Thompson Sampling for graph-based bandits.
problem Efficiently solving bandit problems with smooth payoffs on graphs.
method SpectralTS algorithm for a graph-based bandit problem with effective dimension d.
result SpectralTS offers a computationally more efficient alternative with regret scaling as d*sqrt(T ln N).
Algorithm maximizes revenue from user choices with contextual information.
problem Maximizing revenue from user choices with contextual preference information.
method Proposes an algorithm that learns from user feedback and achieves a revenue regret of order \( \widetilde{O}(d \sqrt{K T} / L_0 ) \).
result Achieves a revenue regret of order \( \widetilde{O}(d \sqrt{K T} / L_0 ) \) and a lower bound of order \( \Omega(d \sqrt{T}/ L_0) \).
A learner selects subsets of choices for a user who then picks from them, aiming to minimize regret.
problem Optimizing subset selection for user choices in a stochastic setting.
method Introduces a new problem and defines regret, then proposes algorithms with matching upper and lower bounds.
result Upper and lower bounds on expected regret match up to a logarithmic term, demonstrating algorithm efficiency.
Adaptive algorithm improves convergence rate of Langevin dynamics.
problem Improving convergence rate of Langevin dynamics.
method Adaptive non-reversible stochastic gradient Langevin dynamics algorithm.
result Improved convergence rate of the algorithm.
This study examines how learning algorithms affect collective action in machine learning.
problem The impact of collective action on machine learning is limited when not considering the choice of learning algorithms.
method Focuses on distributionally robust optimization and stochastic gradient descent, analyzing their effects on collective success.
result The choice of learning algorithm significantly impacts the effective size and success of a collective in machine learning.
New algorithm improves stability of optimization algorithms by adapting step-size.
problem Optimization algorithms' effectiveness is sensitive to step-size hyperparameters.
method Adapts NGN step-size method with momentum to enhance stability.
result Achieves convergence rate of O(1/√K) without restrictive assumptions.
The paper introduces V(I) to guide algorithm choice and parameter tuning in financial forecasting.
problem Selecting optimal algorithms and tuning parameters for financial time-series forecasting.
method Estimating Shannon's mutual information and using it to define performance bounds.
result Illustrates the value of information for mean-square error minimization in cryptocurrency forecasts.
Signature Isolation Forest removes constraints from FIF by using rough path theory's signature transform.
problem Challenges in FIF's linear inner product and dictionary choices leading to unreliable results.
method Introduces Signature Isolation Forest using rough path theory's signature transform to remove linearity constraints.
result Demonstrates relevance of methods through numerical experiments and real-world applications.
The study estimates how changing words in sentences affects audience perception.
problem Estimating the causal effect of lexical choice on audience perception.
method Two classes of methods: quasi-experimental designs and classification problems.
result Algorithmic estimates align with randomized-control trials and can be transferred across domains.
DMNL bandits optimize assortment choices balancing relevance and diversity.
problem Balancing relevance-driven choice with within-assortment diversity.
method Augments MNL choice probabilities with a submodular diversity function, proposing a white-box UCB-based algorithm.
result Achieves at least a ( 1 − 1 e + 1 ) (1-\frac{1}{e+1}) ( 1 − e + 1 1 ) -approximate regret bound of $ ilde{O}\left(d \sqrt{T/K}
ight)$ . A new method reduces high-dimensional state space for dynamic choice models.
problem Estimation of dynamic discrete choice models is computationally intensive and infeasible in high-dimensional settings.
method Recursive partitioning algorithm to reduce dimensionality of high-dimensional state space.
result Our method reduces estimation bias and makes estimation feasible.
A hierarchical clustering algorithm for data clouds without structure assumptions.
problem Exploring data clouds without making structure assumptions.
method Hierarchical topological clustering algorithm that infers persistence of outliers and clusters of arbitrary shape from data hierarchy.
result The algorithm can provide meaningful clusters in complex datasets.
This paper defines resource-constrained classifier performance and its impact on algorithm choice.
problem Classification tasks in resource-constrained settings where actions are limited.
method Defines resource-constrained classifier performance and discusses gains and lift.
result Gains and lift metrics can lead to different algorithm choices.
New model improves website ranking by considering user choices as a whole.
problem Optimizing content ordering for user clicks in website design.
method Introduced multinomial logit (MNL) choice model to LTR framework, proposing UCB algorithms.
result Proved theoretical bounds on regret for UCB algorithms in both known and unknown position parameter settings.
Gaussians as noise in NCE lead to exponentially bad conditioning, hindering its efficiency.
problem Exponential conditioning of Hessian in NCE with Gaussian noise.
method Using Gaussian as the noise distribution in NCE.
result Gaussian noise in NCE leads to exponentially bad conditioning of the loss Hessian.
We explain an algorithm for finding a boundary link Seifert matrix for a given Alexander polynomial. The algorithm depends on several choices and therefore makes it possible to find non-equivalent Seifert matrices for a given Alexander polynomial.
Flexible nonparametric model for discrete choice analysis.
problem Modeling heterogeneity in discrete choice data without fixed component limits.
method Dirichlet process mixture model with expectation maximisation algorithm.
result Proposed model outperforms latent class MNL and mixed MNL models in both fit and predictive ability.
Paper improves anomaly detection by using non-uniform random choices in isolation forests.
problem Detecting clustered diverse outliers more effectively.
method Comparing different split guiding criteria in isolation forests.
result Non-uniform random choices improve outlier discrimination for certain outlier classes.
The paper tackles scalable simulation of discrete random variables.
problem Simulating discrete random variables with general and varying distributions in a scalable framework.
method Inspired by discrete choice models, the paper introduces parallelized randomness and a single associative operation for simulation.
result Characterization of algorithms for scalable simulation of discrete random variables.
ARFF reduces spectral bias in SGD-trained neural networks.
problem Spectral bias in two-layer neural networks.
method Comparison of SGD and ARFF on spectral bias and robustness.
result ARFF yields a closer to zero spectral bias compared to SGD.
We consider the problem of learning the preferences of a heterogeneous population by observing choices from an assortment of products, ads, or other offerings. Our observation model takes a form common in assortment planning applications: each arriving customer is offered an assortment consisting of a subset of all pos…
Platform learns user preferences to avoid abandonment due to marketing fatigue.
problem Maximizing platform's cumulative payoff while avoiding user abandonment.
method Dynamic sequential choice model to balance exploration and exploitation.
result Proposed algorithm achieves optimal regret bound for online learning.
Algorithm optimizes non-convex functions using dueling comparisons.
problem Optimizing non-convex functions with limited function evaluations.
method COMP-GP-UCB algorithm, leveraging dueling-choice bandits.
result Theoretical guarantee of O ( Φ T ) O(\fracΦ{\sqrt{T}}) O ( T Φ ) on simple regret.