Anand Louis
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Anand Louis has a total of 34 co-authors. Here are the top co-authors by number of shared publications:
- Amit Deshpande
Microsoft Research India
6 shared publications - Sruthi Gorantla
Indian Institute of Science Bengaluru India
5 shared publications - Prajakta Nimbhorkar
Chennai Mathematical Institute, Chennai, India
3 shared publications - Amit Deshpande
Microsoft Research, Bengaluru, India
2 shared publications - Atasi Panda
Indian Institute of Science Chennai Mathematical Institute Indian Institute of Science Prajakta Nimbhorkar Chennai Mathematical Institute
2 shared publications - Rameesh Paul
Indian Institute of Science Bangalore India
2 shared publications
Latest Publications
- Open Access
Group Fair Matchings using Convex Cost Functions
We consider the problem of assigning items to platforms where each item has a utility associated with each of the platforms to which it can be assigned. Each platform has a soft constraint over the total number of items it serves, modeled via a convex cost function. Additionally, items are partitioned into groups, and each platform also incurs group-specific convex cost over the number of items from each group that can be assigned to the platform. These costs promote group fairness by penalizing imbalances, yielding a soft variation of fairness notions introduced in prior work, such as Restricted Dominance and Minority protection. Restricted Dominance enforces upper bounds on group representation, while Minority protection enforces lower bounds. Our approach replaces such hard constraints with cost-based penalties, allowing more flexible trade-offs. Our model also captures Nash Social Welfare kind of objective. The cost of an assignment is the sum of the values of all the cost functions across all the groups and platforms. The objective is to find an assignment that minimizes the cost while achieving a total utility that is at least a user-specified threshold. The main challenge lies in balancing the overall platform cost with group-specific costs, both governed by convex functions, while meeting the utility constraint. We present an efficient polynomial-time approximation algorithm, supported by theoretical guarantees and experimental evaluation. Our algorithm is based on techniques involving linear programming and network flows. We also provide an exact algorithm for a special case with uniform utilities and establish the hardness of the general problem when the groups can intersect arbitrarily.
Open AccessGroup Fair Matchings using Convex Cost Functions
We consider the problem of assigning items to platforms where each item has a utility associated with each of the platforms to which it can be assigned. Each platform has a soft constraint over the total number of items it serves, modeled via a convex cost function. Additionally, items are partitioned into groups, and each platform also incurs group-specific convex cost over the number of items from each group that can be assigned to the platform. These costs promote group fairness by penalizing imbalances, yielding a soft variation of fairness notions introduced in prior work, such as Restricted Dominance and Minority protection. Restricted Dominance enforces upper bounds on group representation, while Minority protection enforces lower bounds. Our approach replaces such hard constraints with cost-based penalties, allowing more flexible trade-offs. Our model also captures Nash Social Welfare kind of objective. The cost of an assignment is the sum of the values of all the cost functions across all the groups and platforms. The objective is to find an assignment that minimizes the cost while achieving a total utility that is at least a user-specified threshold. The main challenge lies in balancing the overall platform cost with group-specific costs, both governed by convex functions, while meeting the utility constraint. We present an efficient polynomial-time approximation algorithm, supported by theoretical guarantees and experimental evaluation. Our algorithm is based on techniques involving linear programming and network flows. We also provide an exact algorithm for a special case with uniform utilities and establish the hardness of the general problem when the groups can intersect arbitrarily.
- Open Access
On Sparsest Cut and Conductance in Directed Polymatroidal Networks
We consider algorithms and spectral bounds for sparsest cut and conductance in directed polymatrodal networks. This is motivated by recent work on submodular hypergraphs \cite{Yoshida19,LiM18,ChenOT23,Veldt23} and previous work on multicommodity flows and cuts in polymatrodial networks \cite{ChekuriKRV15}. We obtain three results. First, we obtain an $O(\sqrt{\log n})$-approximation for sparsest cut and point out how this generalizes the result in \cite{ChenOT23}. Second, we consider the symmetric version of conductance and obtain an $O(\sqrt{OPT \log r})$ approximation where $r$ is the maximum degree and we point out how this generalizes previous work on vertex expansion in graphs. Third, we prove a non-constructive Cheeger like inequality that generalizes previous work on hypergraphs. We provide a unified treatment via line-embeddings which were shown to be effective for submodular cuts in \cite{ChekuriKRV15}.Open AccessOn Sparsest Cut and Conductance in Directed Polymatroidal Networks
We consider algorithms and spectral bounds for sparsest cut and conductance in directed polymatrodal networks. This is motivated by recent work on submodular hypergraphs \cite{Yoshida19,LiM18,ChenOT23,Veldt23} and previous work on multicommodity flows and cuts in polymatrodial networks \cite{ChekuriKRV15}. We obtain three results. First, we obtain an $O(\sqrt{\log n})$-approximation for sparsest cut and point out how this generalizes the result in \cite{ChenOT23}. Second, we consider the symmetric version of conductance and obtain an $O(\sqrt{OPT \log r})$ approximation where $r$ is the maximum degree and we point out how this generalizes previous work on vertex expansion in graphs. Third, we prove a non-constructive Cheeger like inequality that generalizes previous work on hypergraphs. We provide a unified treatment via line-embeddings which were shown to be effective for submodular cuts in \cite{ChekuriKRV15}. - Open Access
- 1 August 2024
- Published by International Joint Conferences on Artificial Intelligence
- Vol. 1, 175-183
Individual Fairness under Group Fairness Constraints in Bipartite Matching - One Framework to Approximate Them All
We study the probabilistic assignment of items to platforms that satisfies both group and individual fairness constraints. Each item belongs to specific groups and has a preference ordering over platforms. Each platform enforces group fairness by limiting the number of items per group that can be assigned to it. There could be multiple optimal solutions that satisfy the group fairness constraints, but this alone ignores item preferences. Our approach explores a `best of both worlds fairness' solution to get a randomized matching, which is ex-ante individually fair and ex-post group-fair. Thus, we seek a `probabilistic individually fair' distribution over `group-fair' matchings where each item has a `high' probability of matching to one of its top choices. This distribution is also ex-ante group-fair. Users can customize fairness constraints to suit their requirements. Our first result is a polynomial-time algorithm that computes a distribution over `group-fair' matchings such that the individual fairness constraints are approximately satisfied and the expected size of a matching is close to OPT. We empirically test this on real-world datasets. We present two additional polynomial-time bi-criteria approximation algorithms that users can choose from to balance group fairness and individual fairness trade-offs. For disjoint groups, we provide an exact polynomial-time algorithm adaptable to additional lower `group fairness' bounds. Extending our model, we encompass `maxmin group fairness,' amplifying underrepresented groups, and `mindom group fairness,' reducing the representation of dominant groups.'
- 1 August 2024
- Published by International Joint Conferences on Artificial Intelligence
- Vol. 1, 175-183
Open AccessIndividual Fairness under Group Fairness Constraints in Bipartite Matching - One Framework to Approximate Them All
We study the probabilistic assignment of items to platforms that satisfies both group and individual fairness constraints. Each item belongs to specific groups and has a preference ordering over platforms. Each platform enforces group fairness by limiting the number of items per group that can be assigned to it. There could be multiple optimal solutions that satisfy the group fairness constraints, but this alone ignores item preferences. Our approach explores a `best of both worlds fairness' solution to get a randomized matching, which is ex-ante individually fair and ex-post group-fair. Thus, we seek a `probabilistic individually fair' distribution over `group-fair' matchings where each item has a `high' probability of matching to one of its top choices. This distribution is also ex-ante group-fair. Users can customize fairness constraints to suit their requirements. Our first result is a polynomial-time algorithm that computes a distribution over `group-fair' matchings such that the individual fairness constraints are approximately satisfied and the expected size of a matching is close to OPT. We empirically test this on real-world datasets. We present two additional polynomial-time bi-criteria approximation algorithms that users can choose from to balance group fairness and individual fairness trade-offs. For disjoint groups, we provide an exact polynomial-time algorithm adaptable to additional lower `group fairness' bounds. Extending our model, we encompass `maxmin group fairness,' amplifying underrepresented groups, and `mindom group fairness,' reducing the representation of dominant groups.'
- Open Access
- 10 July 2024
- Published by Association for Computing Machinery (ACM)
- p. 1525-1534
Optimizing Learning-to-Rank Models for Ex-Post Fair Relevance
Learning-to-rank (LTR) models rank items based on specific features, aiming to maximize ranking utility by prioritizing highly relevant items. However, optimizing only for ranking utility can lead to representational harm and may fail to address implicit bias in relevance scores. Prior studies introduced algorithms to train stochastic ranking models, such as the Plackett-Luce ranking model, that maximize expected ranking utility while achieving fairness in expectation (ex-ante fairness). Still, every sampled ranking may not satisfy group fairness (ex-post fairness). Post-processing methods ensure ex-post fairness; however, the LTR model lacks awareness of this step, creating a mismatch between the objective function the LTR model optimizes and the one it is supposed to optimize. In this paper, we first propose a novel objective where the relevance (or the expected ranking utility) is computed over only those rankings that satisfy given representation constraints for groups of items. We call this the ex-post fair relevance. We then give a framework for training Group-Fair LTR models to maximize our proposed ranking objective. Leveraging an efficient sampler for ex-post group-fair rankings and efficient algorithms to train the Plackett-Luce LTR model, we demonstrate their use in training the Group-Fair Plackett-Luce model in our framework. Experiments on MovieLens and Kiva datasets reveal improved fairness and relevance with our group-fair Plackett-Luce model compared to post-processing. In scenarios with implicit bias, our algorithm generally outperforms existing LTR baselines in both fairness and relevance.
- 10 July 2024
- Published by Association for Computing Machinery (ACM)
- p. 1525-1534
Open AccessOptimizing Learning-to-Rank Models for Ex-Post Fair Relevance
Learning-to-rank (LTR) models rank items based on specific features, aiming to maximize ranking utility by prioritizing highly relevant items. However, optimizing only for ranking utility can lead to representational harm and may fail to address implicit bias in relevance scores. Prior studies introduced algorithms to train stochastic ranking models, such as the Plackett-Luce ranking model, that maximize expected ranking utility while achieving fairness in expectation (ex-ante fairness). Still, every sampled ranking may not satisfy group fairness (ex-post fairness). Post-processing methods ensure ex-post fairness; however, the LTR model lacks awareness of this step, creating a mismatch between the objective function the LTR model optimizes and the one it is supposed to optimize. In this paper, we first propose a novel objective where the relevance (or the expected ranking utility) is computed over only those rankings that satisfy given representation constraints for groups of items. We call this the ex-post fair relevance. We then give a framework for training Group-Fair LTR models to maximize our proposed ranking objective. Leveraging an efficient sampler for ex-post group-fair rankings and efficient algorithms to train the Plackett-Luce LTR model, we demonstrate their use in training the Group-Fair Plackett-Luce model in our framework. Experiments on MovieLens and Kiva datasets reveal improved fairness and relevance with our group-fair Plackett-Luce model compared to post-processing. In scenarios with implicit bias, our algorithm generally outperforms existing LTR baselines in both fairness and relevance.
- Open Access
New Approximation Bounds for Small-Set Vertex Expansion
The vertex expansion of the graph is a fundamental graph parameter. Given a graph $G=(V,E)$ and a parameter $δ\in (0,1/2]$, its $δ$-Small-Set Vertex Expansion (SSVE) is defined as \[ \min_{S : |S| = δ|V|} \frac{|{\partial^V(S)}|}{ \min \{ |S|, |S^c| \} } \] where $\partial^V(S)$ is the vertex boundary of a set $S$. The SSVE~problem, in addition to being of independent interest as a natural graph partitioning problem, is also of interest due to its connections to the Strong Unique Games problem. We give a randomized algorithm running in time $n^{{\sf poly}(1/δ)}$, which outputs a set $S$ of size $Θ(δn)$, having vertex expansion at most \[ \max\left(O(\sqrt{φ^* \log d \log (1/δ)}) , \tilde{O}(d\log^2(1/δ)) \cdot φ^* \right), \] where $d$ is the largest vertex degree of the graph, and $φ^*$ is the optimal $δ$-SSVE. The previous best-known guarantees for this were the bi-criteria bounds of $\tilde{O}(1/δ)\sqrt{φ^* \log d}$ and $\tilde{O}(1/δ)φ^* \sqrt{\log n}$ due to Louis-Makarychev [TOC'16]. Our algorithm uses the basic SDP relaxation of the problem augmented with ${\rm poly}(1/δ)$ rounds of the Lasserre/SoS hierarchy. Our rounding algorithm is a combination of the rounding algorithms of Raghavendra-Tan [SODA'12] and Austrin-Benabbas-Georgiou [SODA'13]. A key component of our analysis is novel Gaussian rounding lemma for hyperedges which might be of independent interest.Open AccessNew Approximation Bounds for Small-Set Vertex Expansion
The vertex expansion of the graph is a fundamental graph parameter. Given a graph $G=(V,E)$ and a parameter $δ\in (0,1/2]$, its $δ$-Small-Set Vertex Expansion (SSVE) is defined as \[ \min_{S : |S| = δ|V|} \frac{|{\partial^V(S)}|}{ \min \{ |S|, |S^c| \} } \] where $\partial^V(S)$ is the vertex boundary of a set $S$. The SSVE~problem, in addition to being of independent interest as a natural graph partitioning problem, is also of interest due to its connections to the Strong Unique Games problem. We give a randomized algorithm running in time $n^{{\sf poly}(1/δ)}$, which outputs a set $S$ of size $Θ(δn)$, having vertex expansion at most \[ \max\left(O(\sqrt{φ^* \log d \log (1/δ)}) , \tilde{O}(d\log^2(1/δ)) \cdot φ^* \right), \] where $d$ is the largest vertex degree of the graph, and $φ^*$ is the optimal $δ$-SSVE. The previous best-known guarantees for this were the bi-criteria bounds of $\tilde{O}(1/δ)\sqrt{φ^* \log d}$ and $\tilde{O}(1/δ)φ^* \sqrt{\log n}$ due to Louis-Makarychev [TOC'16]. Our algorithm uses the basic SDP relaxation of the problem augmented with ${\rm poly}(1/δ)$ rounds of the Lasserre/SoS hierarchy. Our rounding algorithm is a combination of the rounding algorithms of Raghavendra-Tan [SODA'12] and Austrin-Benabbas-Georgiou [SODA'13]. A key component of our analysis is novel Gaussian rounding lemma for hyperedges which might be of independent interest. - Open Access
Sparse Cuts in Hypergraphs from Random Walks on Simplicial Complexes
There are a lot of recent works on generalizing the spectral theory of graphs and graph partitioning to hypergraphs. There have been two broad directions toward this goal. One generalizes the notion of graph conductance to hypergraph conductance [LM16, CLTZ18]. In the second approach one can view a hypergraph as a simplicial complex and study its various topological properties [LM06, MW09, DKW16, PR17] and spectral properties [KM17, DK17, KO18a, KO18b, Opp20]. In this work, we attempt to bridge these two directions of study by relating the spectrum of {\em up-down walks} and {\em swap-walks} on the simplicial complex to hypergraph expansion. In surprising contrast to random-walks on graphs, we show that the spectral gap of swap-walks and up-down walks between level $m$ and $l$ with $1 < m \leq l$ can not be used to infer any bounds on hypergraph conductance. Moreover, we show that the spectral gap of swap-walks between $X(1)$ and $X(k-1)$ can not be used to infer any bounds on hypergraph conductance, whereas we give a Cheeger-like inequality relating the spectral of walks between level $1$ and $l$ for any $l \leq k$ to hypergraph expansion. This is a surprising difference between swaps-walks and up-down walks! Finally, we also give a construction to show that the well-studied notion of {\em link expansion} in simplicial complexes can not be used to bound hypergraph expansion in a Cheeger-like manner.Open AccessSparse Cuts in Hypergraphs from Random Walks on Simplicial Complexes
There are a lot of recent works on generalizing the spectral theory of graphs and graph partitioning to hypergraphs. There have been two broad directions toward this goal. One generalizes the notion of graph conductance to hypergraph conductance [LM16, CLTZ18]. In the second approach one can view a hypergraph as a simplicial complex and study its various topological properties [LM06, MW09, DKW16, PR17] and spectral properties [KM17, DK17, KO18a, KO18b, Opp20]. In this work, we attempt to bridge these two directions of study by relating the spectrum of {\em up-down walks} and {\em swap-walks} on the simplicial complex to hypergraph expansion. In surprising contrast to random-walks on graphs, we show that the spectral gap of swap-walks and up-down walks between level $m$ and $l$ with $1 < m \leq l$ can not be used to infer any bounds on hypergraph conductance. Moreover, we show that the spectral gap of swap-walks between $X(1)$ and $X(k-1)$ can not be used to infer any bounds on hypergraph conductance, whereas we give a Cheeger-like inequality relating the spectral of walks between level $1$ and $l$ for any $l \leq k$ to hypergraph expansion. This is a surprising difference between swaps-walks and up-down walks! Finally, we also give a construction to show that the well-studied notion of {\em link expansion} in simplicial complexes can not be used to bound hypergraph expansion in a Cheeger-like manner. - 17 September 2023
- Published by Springer Nature
- p. 727-742
Socially Fair Center-Based and Linear Subspace Clustering
- 17 September 2023
- Published by Springer Nature
- p. 727-742
Socially Fair Center-Based and Linear Subspace Clustering
- Open Access
Optimizing Group-Fair Plackett-Luce Ranking Models for Relevance and Ex-Post Fairness
In learning-to-rank (LTR), optimizing only the relevance (or the expected ranking utility) can cause representational harm to certain categories of items. Moreover, if there is implicit bias in the relevance scores, LTR models may fail to optimize for true relevance. Previous works have proposed efficient algorithms to train stochastic ranking models that achieve fairness of exposure to the groups ex-ante (or, in expectation), which may not guarantee representation fairness to the groups ex-post, that is, after realizing a ranking from the stochastic ranking model. Typically, ex-post fairness is achieved by post-processing, but previous work does not train stochastic ranking models that are aware of this post-processing. In this paper, we propose a novel objective that maximizes expected relevance only over those rankings that satisfy given representation constraints to ensure ex-post fairness. Building upon recent work on an efficient sampler for ex-post group-fair rankings, we propose a group-fair Plackett-Luce model and show that it can be efficiently optimized for our objective in the LTR framework. Experiments on three real-world datasets show that our group-fair algorithm guarantees fairness alongside usually having better relevance compared to the LTR baselines. In addition, our algorithm also achieves better relevance than post-processing baselines, which also ensures ex-post fairness. Further, when implicit bias is injected into the training data, our algorithm typically outperforms existing LTR baselines in relevance.
Open AccessOptimizing Group-Fair Plackett-Luce Ranking Models for Relevance and Ex-Post Fairness
In learning-to-rank (LTR), optimizing only the relevance (or the expected ranking utility) can cause representational harm to certain categories of items. Moreover, if there is implicit bias in the relevance scores, LTR models may fail to optimize for true relevance. Previous works have proposed efficient algorithms to train stochastic ranking models that achieve fairness of exposure to the groups ex-ante (or, in expectation), which may not guarantee representation fairness to the groups ex-post, that is, after realizing a ranking from the stochastic ranking model. Typically, ex-post fairness is achieved by post-processing, but previous work does not train stochastic ranking models that are aware of this post-processing. In this paper, we propose a novel objective that maximizes expected relevance only over those rankings that satisfy given representation constraints to ensure ex-post fairness. Building upon recent work on an efficient sampler for ex-post group-fair rankings, we propose a group-fair Plackett-Luce model and show that it can be efficiently optimized for our objective in the LTR framework. Experiments on three real-world datasets show that our group-fair algorithm guarantees fairness alongside usually having better relevance compared to the LTR baselines. In addition, our algorithm also achieves better relevance than post-processing baselines, which also ensures ex-post fairness. Further, when implicit bias is injected into the training data, our algorithm typically outperforms existing LTR baselines in relevance.
- Open Access
- 8 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 205-216
Sampling Individually-Fair Rankings that are Always Group Fair
- 8 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 205-216
Open AccessSampling Individually-Fair Rankings that are Always Group Fair
- Open Access
Online Algorithms for Matchings with Proportional Fairness Constraints and Diversity Constraints
Matching problems with group-fairness constraints and diversity constraints have numerous applications such as in allocation problems, committee selection, school choice, etc. Moreover, online matching problems have lots of applications in ad allocations and other e-commerce problems like product recommendation in digital marketing. We study two problems involving assigning {\em items} to {\em platforms}, where items belong to various {\em groups} depending on their attributes; the set of items are available offline and the platforms arrive online. In the first problem, we study online matchings with {\em proportional fairness constraints}. Here, each platform on arrival should either be assigned a set of items in which the fraction of items from each group is within specified bounds or be assigned no items; the goal is to assign items to platforms in order to maximize the number of items assigned to platforms. In the second problem, we study online matchings with {\em diversity constraints}, i.e. for each platform, absolute lower bounds are specified for each group. Each platform on arrival should either be assigned a set of items that satisfy these bounds or be assigned no items; the goal is to maximize the set of platforms that get matched. We study approximation algorithms and hardness results for these problems. The technical core of our proofs is a new connection between these problems and the problem of matchings in hypergraphs. Our experimental evaluation shows the performance of our algorithms on real-world and synthetic datasets exceeds our theoretical guarantees.
Open AccessOnline Algorithms for Matchings with Proportional Fairness Constraints and Diversity Constraints
Matching problems with group-fairness constraints and diversity constraints have numerous applications such as in allocation problems, committee selection, school choice, etc. Moreover, online matching problems have lots of applications in ad allocations and other e-commerce problems like product recommendation in digital marketing. We study two problems involving assigning {\em items} to {\em platforms}, where items belong to various {\em groups} depending on their attributes; the set of items are available offline and the platforms arrive online. In the first problem, we study online matchings with {\em proportional fairness constraints}. Here, each platform on arrival should either be assigned a set of items in which the fraction of items from each group is within specified bounds or be assigned no items; the goal is to assign items to platforms in order to maximize the number of items assigned to platforms. In the second problem, we study online matchings with {\em diversity constraints}, i.e. for each platform, absolute lower bounds are specified for each group. Each platform on arrival should either be assigned a set of items that satisfy these bounds or be assigned no items; the goal is to maximize the set of platforms that get matched. We study approximation algorithms and hardness results for these problems. The technical core of our proofs is a new connection between these problems and the problem of matchings in hypergraphs. Our experimental evaluation shows the performance of our algorithms on real-world and synthetic datasets exceeds our theoretical guarantees.