-
FPT Approximations for Capacitated/Fair Clustering with Outliers
Authors:
Rajni Dabas,
Neelima Gupta,
Tanmay Inamdar
Abstract:
Clustering problems such as $k$-Median, and $k$-Means, are motivated from applications such as location planning, unsupervised learning among others. In such applications, it is important to find the clustering of points that is not ``skewed'' in terms of the number of points, i.e., no cluster should contain too many points. This is modeled by capacity constraints on the sizes of clusters. In an o…
▽ More
Clustering problems such as $k$-Median, and $k$-Means, are motivated from applications such as location planning, unsupervised learning among others. In such applications, it is important to find the clustering of points that is not ``skewed'' in terms of the number of points, i.e., no cluster should contain too many points. This is modeled by capacity constraints on the sizes of clusters. In an orthogonal direction, another important consideration in clustering is how to handle the presence of outliers in the data. Indeed, these clustering problems have been generalized in the literature to separately handle capacity constraints and outliers. To the best of our knowledge, there has been very little work on studying the approximability of clustering problems that can simultaneously handle both capacities and outliers.
We initiate the study of the Capacitated $k$-Median with Outliers (C$k$MO) problem. Here, we want to cluster all except $m$ outlier points into at most $k$ clusters, such that (i) the clusters respect the capacity constraints, and (ii) the cost of clustering, defined as the sum of distances of each non-outlier point to its assigned cluster-center, is minimized.
We design the first constant-factor approximation algorithms for C$k$MO. In particular, our algorithm returns a (3+ε)-approximation for C$k$MO in general metric spaces, and a (1+ε)-approximation in Euclidean spaces of constant dimension, that runs in time in time $f(k, m, ε) \cdot |I_m|^{O(1)}$, where $|I_m|$ denotes the input size. We can also extend these results to a broader class of problems, including Capacitated k-Means/k-Facility Location with Outliers, and Size-Balanced Fair Clustering problems with Outliers. For each of these problems, we obtain an approximation ratio that matches the best known guarantee of the corresponding outlier-free problem.
△ Less
Submitted 2 May, 2023;
originally announced May 2023.
-
Near-optimal Algorithms for Stochastic Online Bin Packing
Authors:
Nikhil Ayyadevara,
Rajni Dabas,
Arindam Khan,
K. V. N. Sreenivas
Abstract:
We study the online bin packing problem under two stochastic settings. In the bin packing problem, we are given n items with sizes in (0,1] and the goal is to pack them into the minimum number of unit-sized bins. First, we study bin packing under the i.i.d. model, where item sizes are sampled independently and identically from a distribution in (0,1]. Both the distribution and the total number of…
▽ More
We study the online bin packing problem under two stochastic settings. In the bin packing problem, we are given n items with sizes in (0,1] and the goal is to pack them into the minimum number of unit-sized bins. First, we study bin packing under the i.i.d. model, where item sizes are sampled independently and identically from a distribution in (0,1]. Both the distribution and the total number of items are unknown. The items arrive one by one and their sizes are revealed upon their arrival and they must be packed immediately and irrevocably in bins of size 1. We provide a simple meta-algorithm that takes an offline $α$-asymptotic approximation algorithm and provides a polynomial-time $(α+ \varepsilon)$-competitive algorithm for online bin packing under the i.i.d. model, where $\varepsilon$>0 is a small constant. Using the AFPTAS for offline bin packing, we thus provide a linear time $(1+\varepsilon)$-competitive algorithm for online bin packing under i.i.d. model, thus settling the problem.
We then study the random-order model, where an adversary specifies the items, but the order of arrival of items is drawn uniformly at random from the set of all permutations of the items. Kenyon's seminal result [SODA'96] showed that the Best-Fit algorithm has a competitive ratio of at most 3/2 in the random-order model, and conjectured the ratio to be around 1.15. However, it has been a long-standing open problem to break the barrier of 3/2 even for special cases. Recently, Albers et al. [Algorithmica'21] showed an improvement to 5/4 competitive ratio in the special case when all the item sizes are greater than 1/3. For this special case, we settle the analysis by showing that Best-Fit has a competitive ratio of 1. We make further progress by breaking the barrier of 3/2 for the 3-Partition problem, a notoriously hard special case of bin packing, where all item sizes lie in (1/4,1/2].
△ Less
Submitted 7 May, 2022;
originally announced May 2022.
-
Constant factor approximations for Lower and Upper bounded Clusterings
Authors:
Neelima Gupta,
Sapna Grover,
Rajni Dabas
Abstract:
Clustering is one of the most fundamental problem in Machine Learning. Researchers in the field often require a lower bound on the size of the clusters to maintain anonymity and upper bound for the ease of analysis. Specifying an optimal cluster size is a problem often faced by scientists. In this paper, we present a framework to obtain constant factor approximations for some prominent clustering…
▽ More
Clustering is one of the most fundamental problem in Machine Learning. Researchers in the field often require a lower bound on the size of the clusters to maintain anonymity and upper bound for the ease of analysis. Specifying an optimal cluster size is a problem often faced by scientists. In this paper, we present a framework to obtain constant factor approximations for some prominent clustering objectives, with lower and upper bounds on cluster size. This enables scientists to give an approximate cluster size by specifying the lower and the upper bounds for it. Our results preserve the lower bounds but may violate the upper bound a little. %{GroverGD21_LBUBFL_Cocoon} to $2$. %namely, $k$ Center (LUkC) and $k$ Median (LUkM) problem. We study the problems when either of the bounds is uniform. We apply our framework to give the first constant factor approximations for LUkM and its generalization, $k$-facility location problem (LUkFL), with $β+1$ factor violation in upper bounds where $β$ is the violation of upper bounds in solutions of upper bounded $k$-median and $k$-facility location problems respectively. We also present a result on LUkC with uniform upper bounds and, its generalization, lower and (uniform) upper bounded $k$ supplier problem (LUkS). The approach also gives a result on lower and upper bounded facility location problem (LUFL), improving upon the upper bound violation of $5/2$ due to Gupta et al.
We also reduce the violation in upper bounds for a special case when the gap between the lower and upper bounds is not too small.
△ Less
Submitted 26 March, 2022;
originally announced March 2022.
-
Locating Charging Stations: Connected, Capacitated and Prize- Collecting
Authors:
Rajni Dabas,
Neelima Gupta
Abstract:
In this paper, we study locating charging station problem as facility location problem and its variants ($k$-Median, $k$-Facility location and $k$-center). We study the connectivity and the capacity constraints in these problem.
Capacity and connectivity constraints have been studied in the literature separately for all these problems. We give first constant factor approximations when both the c…
▽ More
In this paper, we study locating charging station problem as facility location problem and its variants ($k$-Median, $k$-Facility location and $k$-center). We study the connectivity and the capacity constraints in these problem.
Capacity and connectivity constraints have been studied in the literature separately for all these problems. We give first constant factor approximations when both the constraints are present. Extending/modifying the techniques used for connected variants of the problem to include capacities or for capacitated variants of problem to include connectivity is a tedious and challenging task. In this paper, we combine the two constraints by reducing the problem to underlying well studied problems, solving them as black box and combine the obtained solutions. We also, combine the two constraints in the prize-collection set up.
In the prize-collecting set up, the problems are not even studied when one of the constraint is present. We present constant factor approximation for them as well.
△ Less
Submitted 13 February, 2022;
originally announced February 2022.
-
On Variants of Facility Location Problem with Outliers
Authors:
Rajni Dabas,
Neelima Gupta
Abstract:
In this work, we study the extension of two variants of the facility location problem (FL) to make them robust towards a few distantly located clients. First, $k$-facility location problem ($k$FL), a common generalization of FL and $k$ median problems, is a well studied problem in literature. In the second variant, lower bounded facility location (LBFL), we are given a bound on the minimum number…
▽ More
In this work, we study the extension of two variants of the facility location problem (FL) to make them robust towards a few distantly located clients. First, $k$-facility location problem ($k$FL), a common generalization of FL and $k$ median problems, is a well studied problem in literature. In the second variant, lower bounded facility location (LBFL), we are given a bound on the minimum number of clients that an opened facility must serve. Lower bounds are required in many applications like profitability in commerce and load balancing in transportation problem. In both the cases, the cost of the solution may be increased grossly by a few distantly located clients, called the outliers. Thus, in this work, we extend $k$FL and LBFL to make them robust towards the outliers. For $k$FL with outliers ($k$FLO) we present the first (constant) factor approximation violating the cardinality requirement by +1. As a by-product, we also obtain the first approximation for FLO based on LP-rounding. For LBFLO, we present a tri-criteria solution with a trade-off between the violations in lower bounds and the number of outliers. With a violation of $1/2$ in lower bounds, we get a violation of $2$ in outliers.
△ Less
Submitted 1 July, 2021;
originally announced July 2021.
-
First Approximation for Uniform Lower and Upper Bounded Facility Location Problem avoiding violation in Lower Bounds
Authors:
Sapna Grover,
Neelima Gupta,
Rajni Dabas
Abstract:
With growing emphasis on e-commerce marketplace platforms where we have a central platform mediating between the seller and the buyer, it becomes important to keep a check on the availability and profitability of the central store. A store serving too less clients can be non-profitable and a store getting too many orders can lead to bad service to the customers which can be detrimental for the bus…
▽ More
With growing emphasis on e-commerce marketplace platforms where we have a central platform mediating between the seller and the buyer, it becomes important to keep a check on the availability and profitability of the central store. A store serving too less clients can be non-profitable and a store getting too many orders can lead to bad service to the customers which can be detrimental for the business. In this paper, we study the facility location problem(FL) with upper and lower bounds on the number of clients an open facility serves. Constant factor approximations are known for the restricted variants of the problem with only the upper bounds or only the lower bounds. The only work that deals with bounds on both the sides violates both the bounds [8]. In this paper, we present the first (constant factor) approximation for the problem violating the upper bound by a factor of (5/2) without violating the lower bounds when both the lower and the upper bounds are uniform. We first give a tri-criteria (constant factor) approximation violating both the upper and the lower bounds and then get rid of violation in lower bounds by transforming the problem instance to an instance of capacitated facility location problem.
△ Less
Submitted 25 June, 2021; v1 submitted 21 June, 2021;
originally announced June 2021.
-
Uniform Capacitated Facility Location Problems with Penalties/Outliers
Authors:
Rajni Dabas,
Neelima Gupta
Abstract:
In this paper, we present a framework to design approximation algorithms for capacitated facility location problems with penalties/outliers using LP-rounding. Primal-dual technique, which has been particularly successful in dealing with outliers and penalties, has not been very successful in dealing with capacities. On the other hand, no primal-dual solution has been able to break the hardness of…
▽ More
In this paper, we present a framework to design approximation algorithms for capacitated facility location problems with penalties/outliers using LP-rounding. Primal-dual technique, which has been particularly successful in dealing with outliers and penalties, has not been very successful in dealing with capacities. On the other hand, no primal-dual solution has been able to break the hardness of capacitated facility location problem(CFLP). LP-rounding techniques had also not been very successful in dealing with capacities until a recent work by Grover et al. \cite{GroverGKP18}. Their constant factor approximation violating the capacities by a small factor ($1 + ε$) is promising while dealing with capacities.Though LP-rounding has not been very promising while dealing with penalties and outliers, we successfully apply it to deal with them along with capacities. That is, our results are obtained by rounding the solution to the natural LP once again exhibiting the power of LP-rounding technique. Solutions obtained by LP-rounding are easy to integrate with other LP-based algorithms.In this paper, we apply our framework to obtain first constant factor approximations for capacitated facility location problem with outlier (CFLPO) and capacitated $k$-facility location problem with penalty(C$k$FLPP) for hard uniform capacities using LP-rounding. Our solutions incur slight violations in capacities, ($1 + ε$) for the problems without cardinality($k$) constraint and ($2 + ε$) for the problems with the cardinality constraint. For the outlier variant, we also incur a small loss ($1 + ε$) in outliers. Due to the hardness of the underlying problems, the violations are inevitable. Thus we achieve the best possible by rounding the solution of natural LP for these problems. To the best of our knowledge, no results are known for CFLPO and C$k$FLPP.
△ Less
Submitted 18 August, 2021; v1 submitted 13 December, 2020;
originally announced December 2020.