Performative Prediction in a Stateful World

Brown, Gavin; Hod, Shlomi; Kalemaj, Iden

doi:10.48550/arxiv.2011.03885

Cited by 5 publications

(7 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Performative prediction. Prior work on performative prediction has largely studied gradientbased optimization methods [31,29,11,5,30,17,27,26,33,10]. Many of the studied procedures only converge to performatively stable points, that is, points θ that satisfy the fixed-point condition…”

Section: Related Workmentioning

confidence: 99%

“…Example 5.1 (Strategic classification). The model (5) arises in strategic classification [15], where agents strategically manipulate their features in response to a deployed model. Suppose the learner uses a linear predictor f θ (x) = θ T x and the agents incur quadratic cost for changing their original features x to manipulated features x , c(x, x ) = 1 2 (x − x )Λ(x − x ).…”

Section: Regret Minimization For Location Familiesmentioning

confidence: 99%

“…Suppose the learner uses a linear predictor f θ (x) = θ T x and the agents incur quadratic cost for changing their original features x to manipulated features x , c(x, x ) = 1 2 (x − x )Λ(x − x ). Then, the best response of an agent, typically modeled as x BR (θ) = argmax x f θ (x ) − c(x, x ), satisfies the structural assumption (5) with z 0 being the agent's original features and µ * = Λ −1 .…”

Section: Regret Minimization For Location Familiesmentioning

confidence: 99%

See 2 more Smart Citations

Regret Minimization with Performative Feedback

Jagadeesan¹,

Zrnic²,

Mendler-Dünner³

2022

Preprint

View full text Add to dashboard Cite

In performative prediction, the deployment of a predictive model triggers a shift in the data distribution. As these shifts are typically unknown ahead of time, the learner needs to deploy a model to get feedback about the distribution it induces. We study the problem of finding near-optimal models under performativity while maintaining low regret. On the surface, this problem might seem equivalent to a bandit problem. However, it exhibits a fundamentally richer feedback structure that we refer to as performative feedback: after every deployment, the learner receives samples from the shifted distribution rather than only bandit feedback about the reward. Our main contribution is regret bounds that scale only with the complexity of the distribution shifts and not that of the reward function. The key algorithmic idea is careful exploration of the distribution shifts that informs a novel construction of confidence bounds on the risk of unexplored models. The construction only relies on smoothness of the shifts and does not assume convexity. More broadly, our work establishes a conceptual approach for leveraging tools from the bandits literature for the purpose of regret minimization with performative feedback.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Regret Minimization For Location Familiesmentioning

confidence: 99%

Section: Regret Minimization For Location Familiesmentioning

confidence: 99%

See 1 more Smart Citation

Regret Minimization with Performative Feedback

Jagadeesan¹,

Zrnic²,

Mendler-Dünner³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…The deployment of learning algorithms in real-world scenarios necessitates versatile and robust algorithms that operate efficiently under mild information structures. Recently, min-max optimization has emerged as a promising framework for framing problems of algorithmic robustness against adversaries [15,31,54], strategically generated data [7,9], and distributional shifts in dynamic environments [59].…”

Section: Introductionmentioning

confidence: 99%

“…One of the primary problem areas in which such an algorithm becomes necessary is in learning from strategically generated or decision-dependent data, a classical problem in operations research (see, e.g., [21] and references therein). This problem has garnered a lot of attention of late in the machine learning community under the name "performative prediction" [7,34,42] due to the growing recognition that learning algorithms are increasingly dealing with data from strategic agents. In such problems, assuming gradient access to the response map of strategic agents is often too restrictive, and the interplay between agents' strategic responses and classic loss functions in machine learning can often weaken the structure present in the underlying learning problem.…”

Section: Introductionmentioning

confidence: 99%

Zeroth-Order Methods for Convex-Concave Minmax Problems: Applications to Decision-Dependent Risk Minimization

Maheshwari¹,

Chiu²,

Mazumdar³

et al. 2021

Preprint

View full text Add to dashboard Cite

Min-max optimization is emerging as a key framework for analyzing problems of robustness to strategically and adversarially generated data. We propose a random reshuffling-based gradient free Optimistic Gradient Descent-Ascent algorithm for solving convex-concave min-max problems with finite sum structure. We prove that the algorithm enjoys the same convergence rate as that of zeroth-order algorithms for convex minimization problems. We further specialize the algorithm to solve distributionally robust, decision-dependent learning problems, where gradient information is not readily available. Through illustrative simulations, we observe that our proposed approach learns models that are simultaneously robust against adversarial distribution shifts and strategic decisions from the data sources, and outperforms existing methods from the strategic classification literature.Preprint. Under review.

show abstract

Approximate Regions of Attraction in Learning with Decision-Dependent Distributions

Dong

Ratliff

2021

Preprint

View full text Add to dashboard Cite

As data-driven methods are deployed in real-world settings, the processes that generate the observed data will often react to the decisions of the learner. For example, a data source may have some incentive for the algorithm to provide a particular label (e.g. approve a bank loan), and manipulate their features accordingly. Work in strategic classification and decision-dependent distributions seeks to characterize the closed-loop behavior of deploying learning algorithms by explicitly considering the effect of the classifier on the underlying data distribution. More recently, works in performative prediction seek to classify the closed-loop behavior by considering general properties of the mapping from classifier to data distribution, rather than an explicit form. Building on this notion, we analyze repeated risk minimization as the perturbed trajectories of the gradient flows of performative risk minimization. We consider the case where there may be multiple local minimizers of performative risk, motivated by real world situations where the initial conditions may have significant impact on the long-term behavior of the system. As a motivating example, we consider a company whose current employee demographics affect the applicant pool they interview: the initial demographics of the company can affect the long-term hiring policies of the company. We provide sufficient conditions to characterize the region of attraction for the various equilibria in this settings. Additionally, we introduce the notion of performative alignment, which provides a geometric condition on the convergence of repeated risk minimization to performative risk minimizers.Preprint. Under review.

show abstract

Performative Prediction in a Stateful World

Cited by 5 publications

References 3 publications

Regret Minimization with Performative Feedback

Regret Minimization with Performative Feedback

Zeroth-Order Methods for Convex-Concave Minmax Problems: Applications to Decision-Dependent Risk Minimization

Approximate Regions of Attraction in Learning with Decision-Dependent Distributions

Contact Info

Product

Resources

About