arXiv subjects

Hanrui Zhang

Publications and source records attributed to Hanrui Zhang.

30 records · Page 2Linked to original sources

arXiv2021

Automated Mechanism Design for Classification with Partial Verification

We study the problem of automated mechanism design with partial verification, where each type can (mis)report only a restricted set of types (rather than any other type), induced by the principal's limited verification power. We prove hardness results when the revelation principle does not necessarily hold, as well as when types have even minimally different preferences. In light of these hardness results, we focus on truthful mechanisms in the setting where all types share the same preference over outcomes, which is motivated by applications in, e.g., strategic classification. We present a number of algorithmic and structural results, including an efficient algorithm for finding optimal deterministic truthful mechanisms, which also implies a faster algorithm for finding optimal randomized truthful mechanisms via a characterization based on convexity. We then consider a more general setting, where the principal's cost is a function of the combination of outcomes assigned to each type. In particular, we focus on the case where the cost function is submodular, and give generalizations of essentially all our results in the classical setting where the cost function is additive. Our results provide a relatively complete picture for automated mechanism design with partial verification.

cs.GT↗

arXiv2021

Classification with Strategically Withheld Data

Machine learning techniques can be useful in applications such as credit approval and college admission. However, to be classified more favorably in such contexts, an agent may decide to strategically withhold some of her features, such as bad test scores. This is a missing data problem with a twist: which data is missing {\em depends on the chosen classifier}, because the specific classifier is what may create the incentive to withhold certain feature values. We address the problem of training classifiers that are robust to this behavior. We design three classification methods: {\sc Mincut}, {\sc Hill-Climbing} ({\sc HC}) and Incentive-Compatible Logistic Regression ({\sc IC-LR}). We show that {\sc Mincut} is optimal when the true distribution of data is fully known. However, it can produce complex decision boundaries, and hence be prone to overfitting in some cases. Based on a characterization of truthful classifiers (i.e., those that give no incentive to strategically hide features), we devise a simpler alternative called {\sc HC} which consists of a hierarchical ensemble of out-of-the-box classifiers, trained using a specialized hill-climbing procedure which we show to be convergent. For several reasons, {\sc Mincut} and {\sc HC} are not effective in utilizing a large number of complementarily informative features. To this end, we present {\sc IC-LR}, a modification of Logistic Regression that removes the incentive to strategically drop features. We also show that our algorithms perform well in experiments on real-world data sets, and present insights into their relative performance in different settings.

cs.LG↗

arXiv2020

On the Equivalence Between High-Order Network-Influence Frameworks: General-Threshold, Hypergraph-Triggering, and Logic-Triggering Models

In this paper, we study several high-order network-influence-propagation frameworks and their connection to the classical network diffusion frameworks such as the triggering model and the general threshold model. In one framework, we use hyperedges to represent many-to-one influence -- the collective influence of a group of nodes on another node -- and define the hypergraph triggering model as a natural extension to the classical triggering model. In another framework, we use monotone Boolean functions to capture the diverse logic underlying many-to-one influence behaviors, and extend the triggering model to the Boolean-function triggering model. We prove that the Boolean-function triggering model, even with refined details of influence logic, is equivalent to the hypergraph triggering model, and both are equivalent to the general threshold model. Moreover, the general threshold model is optimal in the number of parameters, among all models with the same expressive power. We further extend these three equivalent models by introducing correlations among influence propagations on different nodes. Surprisingly, we discover that while the correlated hypergraph-based model is still equivalent to the correlated Boolean-function-based model, the correlated general threshold model is more restrictive than the two high-order models. Our study sheds light on high-order network-influence propagations by providing new insight into the group influence behaviors in existing models, as well as diverse modeling tools for understanding influence propagations in networks.

cs.SI↗

arXiv2020

Mitigating Manipulation in Peer Review via Randomized Reviewer Assignments

We consider three important challenges in conference peer review: (i) reviewers maliciously attempting to get assigned to certain papers to provide positive reviews, possibly as part of quid-pro-quo arrangements with the authors; (ii) "torpedo reviewing," where reviewers deliberately attempt to get assigned to certain papers that they dislike in order to reject them; (iii) reviewer de-anonymization on release of the similarities and the reviewer-assignment code. On the conceptual front, we identify connections between these three problems and present a framework that brings all these challenges under a common umbrella. We then present a (randomized) algorithm for reviewer assignment that can optimally solve the reviewer-assignment problem under any given constraints on the probability of assignment for any reviewer-paper pair. We further consider the problem of restricting the joint probability that certain suspect pairs of reviewers are assigned to certain papers, and show that this problem is NP-hard for arbitrary constraints on these joint probabilities but efficiently solvable for a practical special case. Finally, we experimentally evaluate our algorithms on datasets from past conferences, where we observe that they can limit the chance that any malicious reviewer gets assigned to their desired paper to 50% while producing assignments with over 90% of the total optimal similarity. Our algorithms still achieve this similarity while also preventing reviewers with close associations from being assigned to the same paper.

cs.AI↗

arXiv2020

Planning with Submodular Objective Functions

We study planning with submodular objective functions, where instead of maximizing the cumulative reward, the goal is to maximize the objective value induced by a submodular function. Our framework subsumes standard planning and submodular maximization with cardinality constraints as special cases, and thus many practical applications can be naturally formulated within our framework. Based on the notion of multilinear extension, we propose a novel and theoretically principled algorithmic framework for planning with submodular objective functions, which recovers classical algorithms when applied to the two special cases mentioned above. Empirically, our approach significantly outperforms baseline algorithms on synthetic environments and navigation tasks.

cs.AI↗

arXiv2020

Nearly Linear Row Sampling Algorithm for Quantile Regression

We give a row sampling algorithm for the quantile loss function with sample complexity nearly linear in the dimensionality of the data, improving upon the previous best algorithm whose sampling complexity has at least cubic dependence on the dimensionality. Based upon our row sampling algorithm, we give the fastest known algorithm for quantile regression and a graph sparsification algorithm for balanced directed graphs. Our main technical contribution is to show that Lewis weights sampling, which has been used in row sampling algorithms for $\ell_p$ norms, can also be applied in row sampling algorithms for a variety of loss functions. We complement our theoretical results by experiments to demonstrate the practicality of our approach.

cs.DS↗

arXiv2019

Provably Efficient $Q$-learning with Function Approximation via Distribution Shift Error Checking Oracle

$Q$-learning with function approximation is one of the most popular methods in reinforcement learning. Though the idea of using function approximation was proposed at least 60 years ago, even in the simplest setup, i.e, approximating $Q$-functions with linear functions, it is still an open problem on how to design a provably efficient algorithm that learns a near-optimal policy. The key challenges are how to efficiently explore the state space and how to decide when to stop exploring in conjunction with the function approximation scheme. The current paper presents a provably efficient algorithm for $Q$-learning with linear function approximation. Under certain regularity assumptions, our algorithm, Difference Maximization $Q$-learning (DMQ), combined with linear function approximation, returns a near-optimal policy using a polynomial number of trajectories. Our algorithm introduces a new notion, the Distribution Shift Error Checking (DSEC) oracle. This oracle tests whether there exists a function in the function class that predicts well on a distribution $\mathcal{D}_1$, but predicts poorly on another distribution $\mathcal{D}_2$, where $\mathcal{D}_1$ and $\mathcal{D}_2$ are distributions over states induced by two different exploration policies. For the linear function class, this oracle is equivalent to solving a top eigenvalue problem. We believe our algorithmic insights, especially the DSEC oracle, are also useful in designing and analyzing reinforcement learning algorithms with general function approximation.

cs.LG↗

arXiv2018

Capturing Complementarity in Set Functions by Going Beyond Submodularity/Subadditivity

We introduce two new "degree of complementarity" measures, which we refer to, respectively, as supermodular width and superadditive width. Both are formulated based on natural witnesses of complementarity. We show that both measures are robust by proving that they, respectively, characterize the gap of monotone set functions from being submodular and subadditive. Thus, they define two new hierarchies over monotone set functions, which we will refer to as Supermodular Width (SMW) hierarchy and Superadditive Width (SAW) hierarchy, with level 0 of the hierarchies resting exactly on submodular and subadditive functions, respectively. We present a comprehensive comparative analysis of the SMW hierarchy and the Supermodular Degree (SD) hierarchy, defined by Feige and Izsak. We prove that the SMW hierarchy is strictly more expressive than the SD hierarchy. We show that previous results regarding approximation guarantees for welfare and constrained maximization as well as regarding the Price of Anarchy (PoA) of simple auctions can be extended without any loss from the SD hierarchy to the SMW hierarchy. We also establish almost matching information-theoretical lower bounds. The combination of these approximation and hardness results illustrate that the SMW hierarchy provides an accurate characterization of "near submodularity" needed for maximization approximation. While SD and SMW hierarchies support nontrivial bounds on the PoA of simple auctions, we show that our SAW hierarchy seems to capture more intrinsic properties needed to realize the efficiency of simple auctions. So far, the SAW hierarchy provides the best dependency for the PoA of Single-bid Auction, and is nearly as competitive as the Maximum over Positive Hypergraphs (MPH) hierarchy for Simultaneous Item First Price Auction (SIA). We provide almost tight lower bounds for the PoA of both auctions with respect to the SAW hierarchy.

cs.DS↗

arXiv2018

Complete Submodularity Characterization in the Comparative Independent Cascade Model

We study the propagation of comparative ideas or items in social networks. A full characterization for submodularity in the comparative independent cascade (Com-IC) model of two-idea cascade is given, for competing ideas and complementary ideas respectively, with or without reconsideration. We further introduce One-Shot model where agents show less patience toward ideas, and show that in One-Shot model, only the strongest idea spreads with submodularity.

cs.SI↗

arXiv2018

A Systematic Framework and Characterization of Influence-Based Network Centrality

In this paper, we present a framework for studying the following fundamental question in network analysis: How should one assess the centralities of nodes in an information/influence propagation process over a social network? Our framework systematically extends a family of classical graph-theoretical centrality formulations, including degree centrality, harmonic centrality, and their "sphere-of-influence" generalizations, to influence-based network centralities. We further extend natural group centralities from graph models to influence models, since group cooperation is essential in social influences. This in turn enables us to assess individuals' centralities in group influence settings by applying the concept of Shapley value from cooperative game theory. Mathematically, using the property that these centrality formulations are Bayesian, we prove the following characterization theorem: Every influence-based centrality formulation in this family is the unique Bayesian centrality that conforms with its corresponding graph-theoretical centrality formulation. Moreover, the uniqueness is fully determined by the centrality formulation on the class of layered graphs, which is derived from a beautiful algebraic structure of influence instances modeled by cascading sequences. Our main mathematical result that layered graphs in fact form a basis for the space of influence-cascading-sequence profiles could also be useful in other studies of network influences. We further provide an algorithmic framework for efficient approximation of these influence-based centrality measures. Our study provides a systematic road map for comparative analyses of different influence-based centrality formulations, as well as for transferring graph-theoretical concepts to influence models.

cs.SI↗

arXiv2018

An Improved Algorithm for Incremental DFS Tree in Undirected Graphs

Depth first search (DFS) tree is one of the most well-known data structures for designing efficient graph algorithms. Given an undirected graph $G=(V,E)$ with $n$ vertices and $m$ edges, the textbook algorithm takes $O(n+m)$ time to construct a DFS tree. In this paper, we study the problem of maintaining a DFS tree when the graph is undergoing incremental updates. Formally, we show: Given an arbitrary online sequence of edge or vertex insertions, there is an algorithm that reports a DFS tree in $O(n)$ worst case time per operation, and requires $O\left(\min\{m \log n, n^2\}\right)$ preprocessing time. Our result improves the previous $O(n \log^3 n)$ worst case update time algorithm by Baswana et al. and the $O(n \log n)$ time by Nakamura and Sadakane, and matches the trivial $Ω(n)$ lower bound when it is required to explicitly output a DFS tree. Our result builds on the framework introduced in the breakthrough work by Baswana et al., together with a novel use of a tree-partition lemma by Duan and Zhan, and the celebrated fractional cascading technique by Chazelle and Guibas.

cs.DS↗

arXiv2015

Unit-sphere games

This paper introduces a class of games, called unit-sphere games, where strategies are real vectors with unit 2-norms (or, on a unit-sphere). As a result, they can no longer be interpreted as probability distributions over actions, but rather be thought of as allocations of one unit of resource to actions and the multiplicative payoff effect on each action is proportional to square-root of the amount of resource allocated to that action. The new definition generates a number of interesting consequences. We first characterize sufficient and necessary conditions under which a two-player unit-sphere game has a Nash equilibrium. The characterization effectively reduces solving a unit-sphere game to finding all eigenvalues and eigenvectors of the product of individual payoff matrices. For any unit-sphere game with non-negative payoff matrices, there always exists a unique Nash equilibrium; furthermore, the unique equilibrium is efficiently reachable via Cournot adjustment. In addition, we show that any equilibrium in positive unit-sphere games corresponds to approximate equilibria in the corresponding normal-form games. Analogous but weaker results are extended to positive n-player unit-sphere games.

cs.GT↗