The Johnson-Lindenstrauss transform is a fundamental method for dimension reduction in Euclidean spaces, that can map any dataset of $n$ points into dimension $O(\log n)$ with low distortion of their distances. This dimension bound is tight in general, but one can bypass it for specific problems. Indeed, tremendous progress has been made for clustering problems, especially in the \emph{continuous} setting where centers can be picked from the ambient space $\mathbb{R}^d$. Most notably, for $k$-median and $k$-means, the dimension bound was improved to $O(\log k)$ [Makarychev, Makarychev and Razenshteyn, STOC 2019]. We explore dimension reduction for clustering in the \emph{discrete} setting, where centers can only be picked from the dataset, and present two results that are both parameterized by the doubling dimension of the dataset, denoted as $\operatorname{ddim}$. The first result shows that dimension $O_ε(\operatorname{ddim} + \log k + \log\log n)$ suffices, and is moreover tight, to guarantee that the cost is preserved within factor $1\pmε$ for every set of centers. Our second result eliminates the $\log\log n$ term in the dimension through a relaxation of the guarantee (namely, preserving the cost only for all approximately-optimal sets of centers), which maintains its usefulness for downstream applications. Overall, we achieve strong dimension reduction in the discrete setting, and find that it differs from the continuous setting not only in the dimension bound, which depends on the doubling dimension, but also in the guarantees beyond preserving the optimal value, such as which clusterings are preserved.

Tree embedding has been a fundamental method in algorithm design with wide applications. We focus on the efficiency of building tree embedding in various computational settings under high-dimensional Euclidean \(\mathbb{R}^d\). We devise a new tree embedding construction framework that operates on an arbitrary metric decomposition with bounded diameter, offering a tradeoff between distortion and the locality of its algorithmic steps. This framework works for general metric spaces and may be of independent interest beyond the Euclidean setting. Using this framework, we obtain a dynamic algorithm that maintains an \(O_{\epsilon}(\log n)\)-distortion tree embedding with update time \(\tilde{O}(n^{\epsilon} + d)\) subject to point insertions/deletions, and a massively parallel algorithm that achieves \(O_{\epsilon}(\log n)\)-distortion in \(O(1)\) rounds and total space \(\tilde{O}(n^{1+\epsilon})\) (for constant \(\epsilon \in (0,1)\)). These new tree embedding results allow for a wide range of applications. Notably, under a similar performance guarantee as in our tree embedding algorithms, i.e., \(\tilde{O}(n^{\epsilon} + d)\) update time and \(O(1)\) rounds, we obtain \(O_{\epsilon}(\log n)\)-approximate dynamic and MPC algorithms for \(k\)-median and earth-mover distance in \(\mathbb{R}^d\).

We propose the first local search algorithm for Euclidean clustering that attains an \(O(1)\)-approximation in almost-linear time. Specifically, for Euclidean \(k\)-Means, our algorithm achieves an \(O(c)\)-approximation in \(\tilde O(n^{1+1/c})\) time, for any constant \(c \ge 1\), maintaining the same running time as the previous (non-local-search-based) approach [la Tour and Saulpic, arXiv’2407.11217] while improving the approximation factor from \(O(c^6)\) to \(O(c)\). The algorithm generalizes to any metric space with sparse spanners, delivering efficient constant approximation in \(\ell_p\) metrics, doubling metrics, Jaccard metrics, etc.

The $k$-center problem is a fundamental optimization problem with numerous applications in machine learning, data analysis, data mining, and communication networks. The $k$-center problem has been extensively studied in the classical sequential setting for several decades, and more recently there have been some efforts in understanding the problem in parallel computing, on the Massively Parallel Computation (MPC) model. For now, we have a good understanding of $k$-center in the case where each local MPC machine has sufficient local memory to store some representatives from each cluster, that is, when one has $Ω(k)$ local memory per machine. While this setting covers the case of small values of $k$, for a large number of clusters these algorithms require undesirably large local memory, making them poorly scalable. The case of large $k$ has been considered only recently for the fully scalable low-local-memory MPC model for the Euclidean instances of the $k$-center problem. However, the earlier works have been considering only the constant dimensional Euclidean space, required a super-constant number of rounds, and produced only $k(1+o(1))$ centers whose cost is a super-constant approximation of $k$-center. In this work, we significantly improve upon the earlier results for the $k$-center problem for the fully scalable low-local-memory MPC model. In the low dimensional Euclidean case in $\mathbb{R}^d$, we present the first constant-round fully scalable MPC algorithm for $(2+\varepsilon)$-approximation. We push the ratio further to $(1 + \varepsilon)$-approximation albeit using slightly more $(1 + \varepsilon)k$ centers. All these results naturally extends to slightly super-constant values of $d$. In the high-dimensional regime, we provide the first fully scalable MPC algorithm that in a constant number of rounds achieves an $O(\log n/ \log \log n)$-approximation for $k$-center.

We devise $ε$-coresets for robust $(k,z)$-Clustering with $m$ outliers through black-box reductions to vanilla case. Given an $ε$-coreset construction for vanilla clustering with size $N$, we construct coresets of size $N\cdot \mathrm{poly}\log(kmε^{-1}) + O_z\left(\min\{kmε^{-1}, mε^{-2z}\log^z(kmε^{-1}) \}\right)$ for various metric spaces, where $O_z$ hides $2^{O(z\log z)}$ factors. This increases the size of the vanilla coreset by a small multiplicative factor of $\mathrm{poly}\log(kmε^{-1})$, and the additive term is up to a $(ε^{-1}\log (km))^{O(z)}$ factor to the size of the optimal robust coreset. Plugging in vanilla coreset results of [Cohen-Addad et al., STOC'21], we obtain the first coresets for $(k,z)$-Clustering with $m$ outliers with size near-linear in $k$ while previous results have size at least $Ω(k^2)$ [Huang et al., ICLR'23; Huang et al., SODA'25]. Technically, we establish two conditions under which a vanilla coreset is as well a robust coreset. The first condition requires the dataset to satisfy special structures - it can be broken into "dense" parts with bounded diameter. We combine this with a new bounded-diameter decomposition that has only $O_z(km ε^{-1})$ non-dense points to obtain the $O_z(km ε^{-1})$ additive bound. Another condition requires the vanilla coreset to possess an extra size-preserving property. We further give a black-box reduction that turns a vanilla coreset to the one satisfying the said size-preserving property, leading to the alternative $O_z(mε^{-2z}\log^{z}(kmε^{-1}))$ additive bound. We also implement our reductions in the dynamic streaming setting and obtain the first streaming algorithms for $k$-Median and $k$-Means with $m$ outliers, using space $\tilde{O}(k+m)\cdot\mathrm{poly}(dε^{-1}\logΔ)$ for inputs on the grid $[Δ]^d$.

We give online algorithms for $k$-Means(more generally, $(k, z)$-Clustering) with nearly optimal consistency (a notion suggested by Lattanzi & Vassilvitskii (2017)). Our result turns any $\alpha$-approximate offline algorithm for clustering into an $(1+\epsilon)\alpha^2$-competitive online algorithm for clustering with $O(k \text{poly} \log n)$ consistency. This consistency bound is optimal up to $\text{poly} \log(n)$ factors. Plugging in the offline algorithm that returns the exact optimal solution, we obtain the first $(1 + \epsilon)$-competitive online algorithm for clustering that achieves a linear in $k$ consistency. This simultaneously improves several previous results (Lattanzi & Vassilvitskii, 2017; Fichtenberger et al., 2021). We validate the performance of our algorithm on real datasets by plugging in the practically efficient $k$-Means++ algorithm. Our online algorithm makes $k$-Means++ achieve good consistency with little overhead to the quality of solutions.

We study streaming algorithms for proportionally fair clustering, a notion originally suggested by Chierichetti et al. (2017), in the sliding window model. We show that although there exist efficient streaming algorithms in the insertion-only model, surprisingly no algorithm can achieve finite ratio without violating the fairness constraint in sliding window. Hence, the problem of fair clustering is a rare separation between the insertion-only streaming model and the sliding window model. On the other hand, we show that if the fairness constraint is relaxed by a multiplicative $(1+\varepsilon)$ factor, there exists a $(1 + \varepsilon)$-approximate sliding window algorithm that uses $\text{poly}(k\varepsilon^{-1}\log n)$ space. This achieves essentially the best parameters (up to degree in the polynomial) provided the aforementioned lower bound. We also implement a number of empirical evaluations on real datasets to complement our theoretical results.

We study fair clustering problems in a setting where distance information is obtained from two sources: a strong oracle providing exact distances, but at a high cost, and a weak oracle providing potentially inaccurate distance estimates at a low cost. The goal is to produce a near-optimal fair clustering on $n$ input points with a minimum number of strong oracle queries. This models the increasingly common trade-off between accurate but expensive similarity measures (e.g., large-scale embeddings) and cheaper but inaccurate alternatives. The study of fair clustering in the model is motivated by the important quest of achieving fairness with the presence of inaccurate information. We achieve the first $(1+\varepsilon)$-coresets for fair $k$-median clustering using $\text{poly}\left(\frac{k}{\varepsilon}\cdot\log n\right)$ queries to the strong oracle. Furthermore, our results imply coresets for the standard setting (without fairness constraints), and we could in fact obtain $(1+\varepsilon)$-coresets for $(k,z)$-clustering for general $z=O(1)$ with a similar number of strong oracle queries. In contrast, previous results achieved a constant-factor $(>10)$ approximation for the standard $k$-clustering problems, and no previous work considered the fair $k$-median clustering problem.

We study efficient algorithms for the Euclidean $k$-Center problem, focusing on the regime of large $k$. We take the approach of data reduction by considering $\alpha$-coreset, which is a small subset $S$ of the dataset $P$ such that any $\beta$-approximation on $S$ is an $(\alpha + \beta)$-approximation on $P$. We give efficient algorithms to construct coresets whose size is $k \cdot o(n)$, which immediately speeds up existing approximation algorithms. Notably, we obtain a near-linear time $O(1)$-approximation when $k = n^c$ for any $0 < c < 1$. We validate the performance of our coresets on real-world datasets with large $k$, and we observe that the coreset speeds up the well-known Gonzalez algorithm by up to $4$ times, while still achieving similar clustering cost. Technically, one of our coreset results is based on a new efficient construction of consistent hashing with competitive parameters. This general tool may be of independent interest for algorithm design in high dimensional Euclidean spaces.

We study the problem of computing Chamfer distance in the fully dynamic setting, where two sets of points $A, B \subset \mathbb{R}^{d}$, each of size up to $n$, dynamically evolve through point insertions or deletions and the goal is to efficiently maintain an approximation to $dist_{\mathrm{CH}}(A,B) = \sum_{a \in A} \min_{b \in B} dist(a,b)$, where $dist$ is a distance measure. Chamfer distance is a widely used dissimilarity metric for point clouds, with many practical applications that require repeated evaluation on dynamically changing datasets, e.g., when used as a loss function in machine learning. In this paper, we present the first dynamic algorithm for maintaining an approximation of the Chamfer distance under the $\ell_p$ norm for $p \in$ {$1,2$}. Our algorithm reduces to approximate nearest neighbor (ANN) search with little overhead. Plugging in standard ANN bounds, we obtain $(1+\epsilon)$-approximation in $\tilde{O}(\epsilon^{-d})$ update time and $O(1/\epsilon)$-approximation in $\tilde{O}(d n^{\epsilon^2} \epsilon^{-4})$ update time. We evaluate our method on real-world datasets and demonstrate that it performs competitively against natural baselines.

In latency-sensitive file caching systems such as Content Delivery Networks (CDNs) and Mobile Edge Computing (MEC), the latency of fetching a missing file to the local cache can be significant. Recent studies have revealed that successive requests for the same missing file before the fetching process completes could still suffer latency (so-called delayed hits). Motivated by the practical scenarios, we study the online general file caching problem with delayed hits and bypassing,i.e., a request may be bypassed and processed directly at the remote data center. The objective is to minimize the total request latency. We present a general reduction that turns a traditional file caching algorithm into one that can handle delayed hits. Based on this reduction, we propose an efficient online file caching algorithm, calledCaLa, with an asymptotically tight competitive ratio as$O(Z \log K)$, whereZis the maximum fetching latency of any file andKis the cache size. Extensive simulations on the production data trace from Google and the Yahoo benchmark illustrate thatCaLacan reduce the latency by up to 8.48% compared with the state-of-the-art schemes dealing with delayed hits without bypassing, and this improvement increases to 26.00% if bypassing is allowed. Furthermore, by upgrading the method for estimating files’ weights inCaLa, we proposeCaLa+, which further reduces the total latency by more than 5%.

The Johnson-Lindenstrauss (JL) Lemma introduced the concept of dimension reduction via a random linear map, which has become a fundamental technique in many computational settings. For a set of $n$ points in $\mathbb{R}^d$ and any fixed $ε>0$, it reduces the dimension $d$ to $O(\log n)$ while preserving, with high probability, all the pairwise Euclidean distances within factor $1+ε$. Perhaps surprisingly, the target dimension can be lower if one only wishes to preserve the optimal value of a certain problem on the pointset, e.g., Euclidean max-cut or $k$-means. However, for some notorious problems, like diameter (aka furthest pair), dimension reduction via the JL map to below $O(\log n)$ does not preserve the optimal value within factor $1+ε$. We propose to focus on another regime, of \emph{moderate dimension reduction}, where a problem's value is preserved within factor $α>1$ using target dimension $\log n / poly(α)$. We establish the viability of this approach and show that the famous $k$-center problem is $α$-approximated when reducing to dimension $O(\tfrac{\log n}{α^2}+\log k)$. Along the way, we address the diameter problem via the special case $k=1$. Our result extends to several important variants of $k$-center (with outliers, capacities, or fairness constraints), and the bound improves further with the input's doubling dimension. While our $poly(α)$-factor improvement in the dimension may seem small, it actually has significant implications for streaming algorithms, and easily yields an algorithm for $k$-center in dynamic geometric streams, that achieves $O(α)$-approximation using space $poly(kdn^{1/α^2})$. This is the first algorithm to beat $O(n)$ space in high dimension $d$, as all previous algorithms require space at least $\exp(d)$. Furthermore, it extends to the $k$-center variants mentioned above.

We design new parallel algorithms for clustering in high-dimensional Euclidean spaces. These algorithms run in the Massively Parallel Computation (MPC) model, and are fully scalable, meaning that the local memory in each machine may be $n^σ$ for arbitrarily small fixed $σ>0$. Importantly, the local memory may be substantially smaller than the number of clusters $k$, yet all our algorithms are fast, i.e., run in $O(1)$ rounds. We first devise a fast MPC algorithm for $O(1)$-approximation of uniform facility location. This is the first fully-scalable MPC algorithm that achieves $O(1)$-approximation for any clustering problem in general geometric setting; previous algorithms only provide $\mathrm{poly}(\log n)$-approximation or apply to restricted inputs, like low dimension or small number of clusters $k$; e.g. [Bhaskara and Wijewardena, ICML'18; Cohen-Addad et al., NeurIPS'21; Cohen-Addad et al., ICML'22]. We then build on this facility location result and devise a fast MPC algorithm that achieves $O(1)$-bicriteria approximation for $k$-Median and for $k$-Means, namely, it computes $(1+\varepsilon)k$ clusters of cost within $O(1/\varepsilon^2)$-factor of the optimum for $k$ clusters. A primary technical tool that we introduce, and may be of independent interest, is a new MPC primitive for geometric aggregation, namely, computing for every data point a statistic of its approximate neighborhood, for statistics like range counting and nearest-neighbor search. Our implementation of this primitive works in high dimension, and is based on consistent hashing (aka sparse partition), a technique that was recently used for streaming algorithms [Czumaj et al., FOCS'22].

We consider a generalized poset sorting problem (GPS), in which we are given a query graph $G = (V, E)$ and an unknown poset $\mathcal{P}(V, \prec)$ that is defined on the same vertex set $V$, and the goal is to make as few queries as possible to edges in $G$ in order to fully recover $\mathcal{P}$, where each query $(u, v)$ returns the relation between $u, v$, i.e., $u \prec v$, $v \prec u$ or $u \not \sim v$. This generalizes both the poset sorting problem [Faigle et al., SICOMP 88] and the generalized sorting problem [Huang et al., FOCS 11]. We give algorithms with $\tilde{O}(n\cdot \mathrm{poly}(k))$ query complexity when $G$ is a complete bipartite graph or $G$ is stochastic under the \ER model, where $k$ is the \emph{width} of the poset, and these generalize [Daskalakis et al., SICOMP 11] which only studies complete graph $G$. Both results are based on a unified framework that reduces the poset sorting to partitioning the vertices with respect to a given pivot element, which may be of independent interest. Our study of GPS also leads to a new $\tilde{O}(n^{1 - 1 / (2W)})$ competitive ratio for the so-called weighted generalized sorting problem where $W$ is the number of distinct weights in the query graph. This problem was considered as an open question in [Charikar et al., JCSS 02], and our result makes important progress as it yields the first nontrivial sublinear ratio for general weighted query graphs (for any bounded $W$). We obtain this via an $\tilde{O}(nk + n^{1.5})$ query complexity algorithm for the case where every edge in $G$ is guaranteed to be comparable in the poset, which generalizes a $\tilde{O}(n^{1.5})$ bound for generalized sorting [Huang et al., FOCS 11].

We study the facility location problem in the dynamic setting, where the goal is to efficiently process an intermixed sequence of point insertions and deletions while maintaining a high quality and stable solution. Although the problem has been studied in the context of general metrics and low-dimensional spaces, much remains unknown concerning dynamic facility location in high dimensional spaces. In this work, we present the first fully dynamic algorithm for facility location in high-dimensional spaces $\mathbb{R}^{d}$. For any $c \geq 1$, our algorithm achieves $O(c)$-approximation, supports point updates in $\tilde{O}(\mathrm{poly}(d)n^{1/c + o(1)})$ amortized time and incurs $O(1)$ amortized recourse. More generally, our result shows that despite the linear-time lower bound on the update time for general metrics, it is possible to achieve sub-linear update times for metric spaces that admit dynamic nearest neighbour oracles. Experiments on real datasets confirm that our algorithm achieves high-quality solutions with low running time, and incurs minimal recourse.

We consider a generalization of the Steiner tree problem, the Steiner forest problem , in the Euclidean plane: the input is a multiset \(X\subseteq{\mathbb{R}}^{2}\) , partitioned into \(k\) color classes \(C_{1},\ldots,C_{k}\subseteq X\) . The goal is to find a minimum-cost Euclidean graph \(G\) such that every color class \(C_{i}\) is connected in \(G\) . We study this Steiner forest problem in the streaming setting, where the stream consists of insertions and deletions of points to \(X\) . Each input point \(x {\in} X\) arrives with its color \(\mathsf{color}(x) {\in} [k]\) , and as usual for dynamic geometric streams, the input is restricted to the discrete grid \(\{1,\ldots,\Delta\}^{2}\) . We design a single-pass streaming algorithm that uses \(\operatorname{poly}(k\cdot\log\Delta)\) space and time, and estimates the cost of an optimal Steiner forest solution within ratio arbitrarily close to the famous Euclidean Steiner ratio \(\alpha_{2}\) (currently \(1.1547\leq\alpha_{2}\leq 1.214\) ). This approximation guarantee matches the state-of-the-art bound for streaming Steiner tree, i.e., when \(k=1\) , and it is a major open question to improve the ratio to \(1+\varepsilon\) even for this special case. Our approach relies on a novel combination of streaming techniques, like sampling and linear sketching, with the classical Arora-style dynamic-programming framework for geometric optimization problems, which usually requires large memory and so far has not been applied in the streaming setting. We complement our streaming algorithm for the Steiner forest problem with simple arguments showing that any finite multiplicative approximation requires \(\Omega(k)\) bits of space.

In Mobile Edge Computing, edge servers have limited storage and computing resources that can only support a small number of functions. Meanwhile, mobile applications are becoming more complex, consisting of multiple dependent tasks, modeled as a Directed Acyclic Graph (DAG). When a request arrives, typically in an online manner with a deadline specified, we need to configure the servers and assign the dependent tasks for efficient processing. This work jointly considers the problem of dependent task placement and scheduling with on-demand function configuration on edge servers, aiming to meet as many deadlines as possible. For a single request, when the configuration on each edge server is fixed, we derive FixDoc to find the optimal task placement and scheduling. When the on-demand function configuration is allowed, we propose GenDoc , a novel approximation algorithm, and analyze its additive error from the optimal theoretically. For multiple requests, we derive OnDoc , an online algorithm easy to deploy in practice. Our extensive experiments show that GenDoc outperforms state-of-the-art baselines in processing 86.14% of these unique applications, and reduces their average completion time by at least 24%. The number of deadlines that OnDoc can satisfy is at least 1.9× that of the baselines.

We study the power of uniform sampling for $k$-Median in various metric spaces. We relate the query complexity for approximating $k$-Median, to a key parameter of the dataset, called the balancedness $\beta \in (0, 1]$ (with $1$ being perfectly balanced). We show that any algorithm must make $\Omega(1 / \beta)$ queries to the point set in order to achieve $O(1)$-approximation for $k$-Median. This particularly implies existing constructions of coresets, a popular data reduction technique, cannot be query-efficient. On the other hand, we show a simple uniform sample of $\mathrm{poly}(k \epsilon^{-1} \beta^{-1})$ points suffices for $(1 + \epsilon)$-approximation for $k$-Median for various metric spaces, which nearly matches the lower bound. We conduct experiments to verify that in many real datasets, the balancedness parameter is usually well bounded, and that the uniform sampling performs consistently well even for the case with moderately large balancedness, which justifies that uniform sampling is indeed a viable approach for solving $k$-Median.

$k$-Clustering in $\mathbb{R}^d$ (e.g., $k$-median and $k$-means) is a fundamental machine learning problem. While near-linear time approximation algorithms were known in the classical setting for a dataset with cardinality $n$, it remains open to find sublinear-time quantum algorithms. We give quantum algorithms that find coresets for $k$-clustering in $\mathbb{R}^d$ with $\tilde{O}(\sqrt{nk}d^{3/2})$ query complexity. Our coreset reduces the input size from $n$ to $\mathrm{poly}(k\epsilon^{-1}d)$, so that existing $\alpha$-approximation algorithms for clustering can run on top of it and yield $(1 + \epsilon)\alpha$-approximation. This eventually yields a quadratic speedup for various $k$-clustering approximation algorithms. We complement our algorithm with a nearly matching lower bound, that any quantum algorithm must make $\Omega(\sqrt{nk})$ queries in order to achieve even $O(1)$-approximation for $k$-clustering.

Max-Cut is a fundamental problem that has been studied extensively in various settings. We design an algorithm for Euclidean Max-Cut, where the input is a set of points in ℝd, in the model of dynamic geometric streams, where the input X ⊆ [Δ]d is presented as a sequence of point insertions and deletions. Previously, Frahling and Sohler [STOC 2005] designed a (1+є)-approximation algorithm for the low-dimensional regime, i.e., it uses space exp(d).

Edge computing systems typically handle a wide variety of applications that exhibit diverse degrees of sensitivity to job latency. Therefore, a multitude of utility functions of the job response time need to be considered by the underlying job dispatching and scheduling mechanism. Nonetheless, previous studies in edge computing mainly focused on optimizing a single utility function across all jobs, e.g., linear, sigmoid, or the hard deadline. In this paper, we design online job dispatching and scheduling strategies in which different jobs can be categorized by different non-increasing utility functions. Our goal is to maximize the total utility of all scheduled jobs. We first prove that no online deterministic algorithm could achieve a competitive ratio better than the lower bound$\Omega \left({\frac {1}{\sqrt {\epsilon }}}\right)$under the$(1+\epsilon)$-speed augmentation model. We proceed to propose an online algorithm, named asO4A, for handling jobs with heterogeneous utilities. We prove thatO4Ais$O\left({\frac {1}{\epsilon ^{2}}}\right)$-competitive. We also design its distributed version, i.e.,DO4A. We implementO4AandDO4Aon an edge computing testbed running deep learning inference jobs. With the production trace from Google Cluster, our experimental and large-scale simulation results indicate thatO4Acan increase the total utility by up to 50% compared with state-of-the-art methods. Besides, the performance loss ofDO4Ais only 2% compared withO4Awith a small communication overhead involved. Moreover, both of our algorithms are robust to estimation errors in job processing time and transmission delay.

Motivated by practical generalizations of the classic k-median and k-means objectives, such as clustering with size constraints, fair clustering, and Wasserstein barycenter, we introduce a meta-theorem for designing coresets for constrained-clustering problems. The meta-theorem reduces the task of coreset construction to one on a bounded number of ring instances with a much-relaxed additive error. This reduction enables us to construct coresets using uniform sampling, in contrast to the widely-used importance sampling, and consequently we can easily handle constrained objectives. Notably and perhaps surprisingly, this simpler sampling scheme can yield coresets whose size is independent of n, the number of input points. Our technique yields smaller coresets, and sometimes the first coresets, for a large number of constrained clustering problems, including capacitated clustering, fair clustering, Euclidean Wasserstein barycenter, clustering in minor-excluded graph, and polygon clustering under Fréchet and Hausdorff distance. Finally, our technique yields also smaller coresets for 1-median in low-dimensional Euclidean spaces, specifically of size $\tilde{O}(\varepsilon^{-15})$ in $\mathbb{R}^{2}$ and $\tilde{O}(\varepsilon^{-16})$ in $\mathbb{R}^{3}$.

In Euclidean Uniform Facility Location, the input is a set of clients in $\mathrm{R}^{d}$ and the goal is to place facilities to serve them, so as to minimize the total cost of opening facilities plus connecting the clients. We study the classical setting of dynamic geometric streams, where the clients are presented as a sequence of insertions and deletions of points in the grid $\{1,ldots\,\Delta \}^{d}$, and we focus on the high-dimensional regime, where the algorithm’s space complexity must be polynomial (and certainly not exponential) in $d \cdot \log \Delta$.We present a new algorithmic framework, based on importance sampling from the stream, for $O(1)$-approximation of the optimal cost using only poly $(d\cdot\log\Delta)$ space. This framework is easy to implement in two passes, one for sampling points and the other for estimating their contribution. Over random-order streams, we can extend this to a one-pass algorithm by using the two halves of the stream separately. Our main result, for arbitrary-order streams, computes $O(d^{1.5})$-approximation in one pass by using the new framework but combining the two passes differently. This improves upon previous algorithms that either need space exponential in d or only guarantee $O(d\cdot\log^{2}\Delta)$-approximation, and therefore our algorithms for high-dimensional streams are the first to avoid the $O(\log\Delta)$ factor in approximation that is inherent to the widely-used quadtree decomposition. Our improvement is achieved by employing a geometric hashing scheme that maps points in $\mathbb{R}^{d}$ into buckets of bounded diameter, with the key property that every point set of small-enough diameter is hashed into at most poly $(d)$ distinct buckets.Finally, we complement our results with a proof that every streaming 1.085-approximation algorithm requires space exponential in poly $(d \cdot log \Delta)$, even for insertion-only streams.

We consider an important generalization of the Steiner tree problem, the Steiner forest problem, in the Euclidean plane: the input is a multiset X ⊆ ℝ², partitioned into k color classes C₁, C₂, …, Cₖ ⊆ X. The goal is to find a minimum-cost Euclidean graph G such that every color class Cᵢ is connected in G. We study this Steiner forest problem in the streaming setting, where the stream consists of insertions and deletions of points to X. Each input point x ∈ X arrives with its color color(x) ∈ [k], and as usual for dynamic geometric streams, the input is restricted to the discrete grid {0, …, Δ}². We design a single-pass streaming algorithm that uses poly(k ⋅ log Δ) space and time, and estimates the cost of an optimal Steiner forest solution within ratio arbitrarily close to the famous Euclidean Steiner ratio α₂ (currently 1.1547 ≤ α₂ ≤ 1.214). This approximation guarantee matches the state of the art bound for streaming Steiner tree, i.e., when k = 1. Our approach relies on a novel combination of streaming techniques, like sampling and linear sketching, with the classical Arora-style dynamic-programming framework for geometric optimization problems, which usually requires large memory and has so far not been applied in the streaming setting. We complement our streaming algorithm for the Steiner forest problem with simple arguments showing that any finite approximation requires Ω(k) bits of space.

In latency-sensitive file caching systems such as Content Delivery Networks (CDNs) and Mobile Edge Computing (MEC), the latency of fetching a missing file to the local cache can be significant. Recent studies have revealed that successive requests of the same missing file before the fetching completes could still suffer latency (so-called delayed hits).Motivated by the practical scenarios, we study the online general file caching problem with delayed hits and bypassing, i.e., a request may be bypassed and processed directly at the remote data center. The objective is to minimize the total request latency. We show a general reduction that turns a traditional file caching algorithm to one that can handle delayed hits. We give an O(Z3/2logK)-competitive algorithm called CaLa with this reduction, where Z is the maximum fetching latency of any file and K is the cache size, and we show a nearly-tight lower bound Ω(Z logK) for our ratio. Extensive simulations based on the production data trace from Google and the Yahoo benchmark illustrate that CaLa can reduce the latency by up to 9.42% compared with the state-of-the-art scheme dealing with delayed hits without bypassing, and this improvement increases to 32.01% if bypassing is allowed.

We provide the first coreset for clustering points in $\mathbb{R}^d$ that have multiple missing values (coordinates). Previous coreset constructions only allow one missing coordinate. The challenge in this setting is that objective functions, like \kMeans, are evaluated only on the set of available (non-missing) coordinates, which varies across points. Recall that an $\epsilon$-coreset of a large dataset is a small proxy, usually a reweighted subset of points, that $(1+\epsilon)$-approximates the clustering objective for every possible center set.Our coresets for $k$-Means and $k$-Median clustering have size $(jk)^{O(\min(j,k))} (\epsilon^{-1} d \log n)^2$, where $n$ is the number of data points, $d$ is the dimension and $j$ is the maximum number of missing coordinates for each data point. We further design an algorithm to construct these coresets in near-linear time, and consequently improve a recent quadratic-time PTAS for $k$-Means with missing values [Eiben et al., SODA 2021] to near-linear time.We validate our coreset construction, which is based on importance sampling and is easy to implement, on various real data sets. Our coreset exhibits a flexible tradeoff between coreset size and accuracy, and generally outperforms the uniform-sampling baseline. Furthermore, it significantly speeds up a Lloyd's-style heuristic for $k$-Means with missing values.

Coresets are modern data-reduction tools that are widely used in data analysis to improve efficiency in terms of running time, space and communication complexity. Our main result is a fast algorithm to construct a small coreset for k-Median in (the shortest-path metric of) an excluded-minor graph. Specifically we give the first coreset of size that depends only on k, ∊ and the excluded-minor size, and our running time is quasi-linear (in the size of the input graph). The main innovation in our new algorithm is that is iterative; it first reduces the n input points to roughly O(log n) reweighted points, then to O(log log n), and so forth until the size is independent of n. Each step in this iterative size reduction is based on the importance sampling framework of Feldman and Langberg (STOC 2011), with a crucial adaptation that reduces the number of distinct points, by employing a terminal embedding (where low distortion is guaranteed only for the distance from every terminal to all other points). Our terminal embedding is technically involved and relies on shortest-path separators, a standard tool in planar and excluded-minor graphs. Furthermore, our new algorithm is applicable also in Euclidean metrics, by simply using a recent terminal embedding result of Narayanan and Nelson (STOC 2019), which extends the Johnson-Lindenstrauss Lemma. We thus obtain an efficient coreset construction in high-dimensional Euclidean spaces, thereby matching and simplifying state-of-the-art results (Sohler and Woodruff, FOCS 2018; Huang and Vishnoi, STOC 2020). In addition, we also employ terminal embedding with additive distortion to obtain small coresets in graphs with bounded highway dimension, and use applications of our coresets to obtain improved approximation schemes, e.g., an improved PTAS for planar k-Median via a new centroid set.

The Bulk Synchronous Parallel (BSP) paradigm is gaining tremendous importance recently due to the popularity of computations as distributed machine learning and graph computation. In a typical BSP job, multiple workers concurrently conduct iterative computations, where frequent synchronization is required. Therefore, the workers should be scheduled simultaneously and their placement on different computing devices could significantly affect the performance. Simply retrofitting a traditional scheduling discipline will likely not yield the desired performance due to the unique characteristics of BSP jobs. In this work, we deriveSPIN, a novel scheduling designed for BSP jobs with placement-sensitive execution to minimize the makespan of all jobs. We first prove the problem approximation hardness and then present howSPINsolves it with a rounding-based randomized approximation approach. Our analysis indicatesSPINachieves a good performance guarantee efficiently. Moreover,SPINis robust against misestimation of job execution time by theoretically bounding its negative impact. We implementSPINon a production-trace driven testbed with 40 GPUs. Our extensive experiments show thatSPINcan reduce the job makespan and the average job completion time by up to$3\times $and$4.68\times $, respectively.SPINalso demonstrates better robustness to execution time misestimation compared with state-of-the-art heuristic baselines.

Motivated by practical scenarios in areas such as Mobile Edge Computing (MEC) and Content Delivery Networks (CDNs), we study online file caching on multiple caches, where a file request might be relayed to other caches or bypassed directly to the memory when a cache miss happens. We can also choose to fetch files from the memory to caches and conduct file replacement if necessary. We take the relaying, bypassing and fetching costs altogether into consideration. We first show the inherent difficulty of the problem even when the online requests are of uniform costs. We propose an O(logK)-competitive randomized online multiple caching algorithm (named Camul) and an O(K)-competitive deterministic algorithm (named Camul-Det), where K is the total number of slots in all caches. Both of them achieve asymptotically optimal competitive ratios. Moreover, our algorithms can be implemented efficiently such that each request is processed in amortized constant time. We conduct extensive simulations on production data traces from Google and a benchmark workload from Yahoo. It shows that our algorithms dramatically outperform state-of-the-art schemes, i.e., reducing the total cost by 85% and 43% respectively compared with important baselines and their strengthened versions with request relaying. More importantly, Camul achieves such a good total cost without sacrificing other performance measures, e.g., the hit ratio, and performs consistently well on various settings of experiment parameters.

We initiate the study of coresets for clustering in graph metrics, i.e., the shortest-path metric of edge-weighted graphs. Such clustering problems are essential to data analysis and used for example in road networks and data visualization. A coreset is a compact summary of the data that approximately preserves the clustering objective for every possible center set, and it offers significant efficiency improvements in terms of running time, storage, and communication, including in streaming and distributed settings. Our main result is a near-linear time construction of a coreset for k-Median in a general graph $G$, with size $O_{\epsilon, k}(\mathrm{tw}(G))$ where $\mathrm{tw}(G)$ is the treewidth of $G$, and we complement the construction with a nearly-tight size lower bound. The construction is based on the framework of Feldman and Langberg [STOC 2011], and our main technical contribution, as required by this framework, is a uniform bound of $O(\mathrm{tw}(G))$ on the shattering dimension under any point weights. We validate our coreset on real-world road networks, and our scalable algorithm constructs tiny coresets with high accuracy, which translates to a massive speedup of existing approximation algorithms such as local search for graph k-Median.

The Bulk Synchronous Parallel (BSP) paradigm is gaining tremendous importance recently because of the pop-ularity of computations such as distributed machine learning and graph computation. In a typical BSP job, multiple workers concurrently conduct iterative computations, where frequent synchronization is required. Therefore, the workers should be scheduled simultaneously and their placement on different computing devices could significantly affect the performance. Simply retrofitting a traditional scheduling discipline will likely not yield the desired performance due to the unique characteristics of BSP jobs. In this work, we derive SPIN, a novel scheduling designed for BSP jobs with placement-sensitive execution to minimize the makespan of all jobs. We first prove the problem approximation hardness and then present how SPIN solves it with a rounding-based randomized approximation approach. Our analysis indicates SPIN achieves a good performance guarantee efficiently. Moreover, SPIN is robust against misestimation of job execution time by theoretically bounding its negative impact. We implement SPIN on a production-trace driven testbed with 40 GPUs. Our extensive experiments show that SPIN can reduce the job makespan and the average job completion time by up to 3× and 4.68×, respectively. Our approach also demonstrates better robustness to execution time misestimation compared with heuristic baselines.

Edge computing systems typically handle a wide variety of applications that exhibit diverse degrees of sensitivity to job latency. Therefore, a multitude of utility functions of the job response time need to be considered by the underlying job dispatching and scheduling mechanism. Nonetheless, previous works in edge computing mainly focused on either one kind of utility function (e.g., linear, sigmoid, or the hard deadline) or different kinds of utilities separately. In this paper, we investigate online job dispatching and scheduling strategies under the setting of coexistence of heterogeneous utilities, i.e., various coexisting jobs can employ different non-increasing utility functions. The goal is to maximize the total utility over all jobs in an edge system. Besides heterogeneous utilities, we here adopt a practical online model where the unrelated machine model and the upload and download delay are considered. We proceed to propose an online algorithm, O4A, to dispatch and schedule jobs with heterogeneous utilities. Our theoretical analysis shows that O4A is O(1/ɛ2)-competitive under the (1 + ɛ)-speed augmentation model, where ɛ is a small positive constant. We implement O4A on an edge computing testbed running deep learning inference jobs. With the production trace from Google Cluster, our experimental and large-scale simulation results indicate that O4A can increase the total utility by up to 39.42% compared with state-of-the-art utility-agnostic methods. Moreover, O4A is robust to estimation errors in job processing time and transmission delay.

We present a unified (randomized) polynomial-time approximation scheme (PTAS) for the prize collecting traveling salesman problem (PCTSP) and the prize collecting Steiner tree problem (PCSTP) in doubling metrics. Given a metric space and a penalty function on a subset of points known as terminals, a solution is a subgraph on points in the metric space whose cost is the weight of its edges plus the penalty due to terminals not covered by the subgraph. Under our unified framework, the solution subgraph needs to be Eulerian for PCTSP, while it needs to be a tree for PCSTP. Before our work, even a QPTAS for the problems in doubling metrics is not known. Our unified PTAS is based on the previous dynamic programming frameworks proposed in Talwar (STOC 2004) and Bartal, Gottlieb, Krauthgamer (STOC 2012). However, since it is unknown which part of the optimal cost is due to edge lengths and which part is due to penalties of uncovered terminals, we need to develop new techniques to apply previous divide-and-conquer strategies and sparse instance decompositions.

We design coresets for Ordered k-Median, a generalization of classical clustering problems such as k-Median and k-Center. Its objective function is defined via the Ordered Weighted Averaging (OWA) paradigm of Yager (1988), where data points are weighted according to a predefined weight vector, but in order of their contribution to the objective (distance from the centers). A powerful data-reduction technique, called a coreset, is to summarize a point set $X$ in $\mathbb{R}^d$ into a small (weighted) point set $X’$, such that for every set of $k$ potential centers, the objective value of the coreset $X’$ approximates that of $X$ within factor $1\pm \epsilon$. When there are multiple objectives (weights), the above standard coreset might have limited usefulness, whereas in a simultaneous coreset, the above approximation holds for all weights (in addition to all centers). Our main result is a construction of a simultaneous coreset of size $O_{\epsilon, d}(k^2 \log^2 |X|)$ for Ordered k-Median. We validate our algorithm on a real geographical data set, and we find our coreset leads to a massive speedup of clustering computations, while maintaining high accuracy for a range of weights.

Motivated by practical scenarios in areas such as Mobile Edge Computing (MEC) and Content Delivery Networks (CDNs), we study online file caching on multiple caches, where a file request might be relayed to other caches or bypassed directly to the memory when a cache miss happens. We take the relaying, bypassing and fetching costs altogether into consideration. We first show the inherent difficulty of the problem even when the online requests are of uniform cost. We propose an O(log K)-competitive randomized algorithm Camul and an O(K)-competitive deterministic algorithm Camul-Det, where K is the total number of slots in all caches. Both online algorithms achieve asymptotically optimal competitive ratios, and can be implemented efficiently such that each request is processed in amortized constant time. We conduct extensive simulations on production data traces from Google and a benchmark workload from Yahoo. It shows that our algorithms dramatically outperform existing schemes, i.e., reducing the total cost by 85% and 43% respectively compared with important baselines and their strengthened versions with request relaying. More importantly, Camul achieves such a good total cost without sacrificing other performance measures, e.g., the hit ratio, and can perform consistently well on various settings of experiment parameters.

In Mobile Edge Computing (MEC), each edge server can be configured with only a small number of functions due to the limited capacity of various resources. Meanwhile, mobile applications become more complicated, consisting of multiple dependent tasks which are typically modeled as a Directed Acyclic Graph (DAG). In edge computing, when an application arrives, we need to place and schedule its tasks onto edge servers and/or the remote cloud, where the functions to execute the tasks are configured. In this work, we jointly consider the problem of dependent task placement and scheduling with on-demand function configuration on servers. Our objective is to minimize the application completion time. Specifically, for the special case when the configuration on each edge server is fixed, we derive an algorithm to find the optimal task placement and scheduling efficiently. When the on-demand function configuration is allowed, we propose a novel approximation algorithm, named GenDoc, and analyze theoretically its additive error from the optimal solution. Our extensive experiments on the cluster trace from Alibaba (including 20365 unique applications with DAG information) show that GenDoc outperforms state-of-the-art baselines in processing 86.14% of these unique applications, and reduces their average completion time by at least 24% (and up to 54%). Moreover, GenDoc consistently performs well on various settings of key parameters.

In a recent work, \cite{chierichetti2017fair} studied the following ``fair'' variants of classical clustering problems such as k-means and k-median: given a set of n data points in R^d and a binary type associated to each data point, the goal is to cluster the points while ensuring that the proportion of each type in each cluster is roughly the same as its underlying proportion. Subsequent work has focused on either extending this setting to when each data point has multiple, non-disjoint sensitive types such as race and gender \cite{bera2019fair}, or to address the problem that the clustering algorithms in the above work do not scale well. The main contribution of this paper is an approach to clustering with fairness constraints that involve {\em multiple, non-disjoint} attributes, that is {\em also scalable}. Our approach is based on novel constructions of coresets: for the k-median objective, we construct an \eps-coreset of size O(\Gamma k^2 \eps^{-d}) where \Gamma is the number of distinct collections of groups that a point may belong to, and for the k-means objective, we show how to construct an \eps-coreset of size O(\Gamma k^3\eps^{-d-1}). The former result is the first known coreset construction for the fair clustering problem with the k-median objective, and the latter result removes the dependence on the size of the full dataset as in~\cite{schmidt2018fair} and generalizes it to multiple, non-disjoint attributes. Importantly, plugging our coresets into existing algorithms for fair clustering such as \cite{backurs2019scalable} results in the fastest algorithms for several cases. Empirically, we assess our approach over the \textbf{Adult} and \textbf{Bank} dataset, and show that the coreset sizes are much smaller than the full dataset; applying coresets indeed accelerates the running time of computing the fair clustering objective while ensuring that the resulting objective difference is small.

In edge-cloud computing, a set of servers (called edge servers) are deployed near the mobile devices to allow these devices to offload their jobs to and subsequently obtain their results from the edge servers with low latency. One fundamental problem in edge-cloud systems is how to dispatch and schedule the jobs so that the job response time (defined as the interval between the release of the job and the arrival of the computation result at the device) is minimized. In this paper, we propose a general model for this problem, where the jobs are generated in arbitrary order and at arbitrary times at the mobile devices and then offloaded to servers with both upload and download delays. Our goal is to minimize the total weighted response time of all the jobs. The weight is set based on how latency-sensitive the job is. We derive the first online job dispatching and scheduling algorithm in edge-clouds, called OnDisc, which is scalable in the speed augmentation model; that is, OnDisc is (1 + ε)-speed O(1/ε)-competitive for any small constant ε > 0. Moreover, OnDisc can be easily implemented in distributed systems. We also extend OnDisc with a fairness knob to incorporate the trade-off between the average job response time and the degree of fairness among jobs. Extensive simulations based on a real-world data-trace from Google show that OnDisc can reduce the total weighted response time dramatically compared with heuristic algorithms.

A coflow is a collection of related parallel flows that occur typically between two stages of a multi-stage computing task in a network, such as shuffle flows in MapReduce. The coflow abstraction allows applications to convey their semantics to the network so that application-level requirements can be better satisfied. In this paper, we study the routing and scheduling of multiple coflows to minimize the total weighted coflow completion time (CCT). We first propose a rounding-based randomized approximation algorithm, called OneCoflow, for single coflow routing and scheduling. The multiple coflow problem is more challenging as coexisting coflows will compete for the same network resources, such as link bandwidth. To minimize the total weighted CCT, we derive an online multiple coflow routing and scheduling algorithm, called OMCoflow. We then derive a competitive ratio bound of our problem and prove that the competitive ratio of OMCoflow is nearly tight. To the best of our knowledge, this is the first online algorithm with theoretical performance guarantees which considers routing and scheduling simultaneously for multi-coflows. Compared with existing methods, OMCoflow runs more efficiently and avoids frequently rerouting the flows. Extensive simulations on a Facebook data trace show that OMCoflow outperforms the state-of-the-art heuristic schemes significantly (e.g., reducing the total weighted CCT by up to 41.8% and the execution time by up to 99.2% against RAPIER).

We study the problem of constructing ε-coresets for the (k, z)-clustering problem in a doubling metric M(X, d). An ε-coreset is a weighted subset S ⊆ X with weight function w : S → ℝ≥0, such that for any k-subset C ∈ [X]k, it holds that Σx∈Sw(x) · dz(x, C) ∈ (1 ± ε) · Σx∈Xdz(x, C). We present an efficient algorithm that constructs an ε-coreset for the (k, z)-clustering problem in M(X, d), where the size of the coreset only depends on the parameters k, z, ε and the doubling dimension ddim(M). To the best of our knowledge, this is the first efficient c-coreset construction of size independent of |X| for general clustering problems in doubling metrics. To this end, we establish the first relation between the doubling dimension of M(X, d) and the shattering dimension (or VC-dimension) of the range space induced by the distance d. Such a relation is not known before, since one can easily construct instances in which neither one can be bounded by (some function of) the other. Surprisingly, we show that if we allow a small (1 ± ε)-distortion of the distance function d (the distorted distance is called the smoothed distance function), the shattering dimension can be upper bounded by O(ε-O(ddim(M))). For the purpose of coreset construction, the above bound does not suffice as it only works for unweighted spaces. Therefore, we introduce the notion of τ-error probabilistic shattering dimension, and prove a (drastically better) upper bound of O(ddim(M)·log(1/ε)+log log 1/τ) for the probabilistic shattering dimension for weighted doubling metrics. As it turns out, an upper bound for the probabilistic shattering dimension is enough for constructing a small coreset. We believe the new relation between doubling and shattering dimensions is of independent interest and may find other applications. Furthermore, we study robust coresets for (k, z)-clustering with outliers in a doubling metric. We show an improved connection between α-approximation and robust coresets. This also leads to improvement upon the previous best known bound of the size of robust coreset for Euclidean space [Feldman and Langberg, STOC 11]. The new bound entails a few new results in clustering and property testing. As another application, we show constant-sized (ε, k, z)centroid sets in doubling metrics can be constructed by extending our coreset construction. Prior to our result, constantsized centroid sets for general clustering problems were only known for Euclidean spaces. We can apply our centroid set to accelerate the local search algorithm (studied in [Friggstad et al., FOCS 2016]) for the (k, z)-clustering problem in doubling metrics.

We study the online submodular maximization problem with free disposal under a matroid constraint. Elements from some ground set arrive one by one in rounds, and the algorithm maintains a feasible set that is independent in the underlying matroid. In each round when a new element arrives, the algorithm may accept the new element into its feasible set and possibly remove elements from it, provided that the resulting set is still independent. The goal is to maximize the value of the final feasible set under some monotone submodular function, to which the algorithm has oracle access. For k -uniform matroids, we give a deterministic algorithm with competitive ratio at least 0.2959, and the ratio approaches 1/α ∞ ≈ 0.3178 as k approaches infinity, improving the previous best ratio of 0.25 by Chakrabarti and Kale (IPCO 2014), Buchbinder et al. (SODA 2015), and Chekuri et al. (ICALP 2015). We also show that our algorithm is optimal among a class of deterministic monotone algorithms that accept a new arriving element only if the objective is strictly increased. Further, we prove that no deterministic monotone algorithm can be strictly better than 0.25-competitive even for partition matroids, the most modest generalization of k -uniform matroids, matching the competitive ratio by Chakrabarti and Kale (IPCO 2014) and Chekuri et al. (ICALP 2015). Interestingly, we show that randomized algorithms are strictly more powerful by giving a (non-monotone) randomized algorithm for partition matroids with ratio 1/α ∞ ≈ 0.3178.

We consider the Traveling Salesman Problem with Neighborhoods (TSPN) in doubling metrics. The goal is to find the shortest tour that visits each of a given collection of subsets (regions or neighborhoods) in the underlying metric space. We give a randomized polynomial-time approximation scheme (PTAS) when the regions are fat weakly disjoint. This notion of regions was first defined when a QPTAS was given for the problem in SODA 2010 (Chan and Elbassioni 2010). The regions are partitioned into a constant number of groups, where in each group, regions should have a common upper bound on their diameters and each region designates one point within it such that these points are far away from one another. We combine the techniques in the previous work, together with the recent PTAS for TSP (STOC 2012: Bartal, Gottlieb, and Krauthgamer 2012) to achieve a PTAS for TSPN. However, several nontrivial technical hurdles need to be overcome for applying the PTAS framework to TSPN: (1) Heuristic to detect sparse instances. In the STOC 2012 paper, a minimum spanning tree heuristic is used to estimate the portion of an optimal tour within some ball. However, for TSPN, it is not known if an optimal tour would use points inside the ball to visit regions that intersect the ball. (2) Partially cut regions in the recursion. After a sparse ball is identified by the heuristic, the PTAS framework for TSP uses dynamic programming to solve the instance restricted to the sparse ball and recurse on the remaining instance. However, for TSPN, it is an important issue to decide whether each region partially intersecting the sparse ball should be solved in the sparse instance or considered in the remaining instance. Surprisingly, we show that both issues can be resolved by conservatively making the ball in question responsible for all intersecting regions. In particular, a sophisticated charging argument is needed to bound the cost of combining tours in the recursion. Moreover, more refined procedures are used to improve the dependence of the running time on the doubling dimension k from the previous exp[( O (1)) k 2 ] (even for just TSP) to exp[2 O ( k log k ) ].

We consider the online vector packing problem in which we have a d dimensional knapsack and items u with weight vectors w_u in R_+^d arrive online in an arbitrary order. Upon the arrival of an item, the algorithm must decide immediately whether to discard or accept the item into the knapsack. When item u is accepted, w_u(i) units of capacity on dimension i will be taken up, for each i in [d]. To satisfy the knapsack constraint, an accepted item can be later disposed of with no cost, but discarded or disposed of items cannot be recovered. The objective is to maximize the utility of the accepted items S at the end of the algorithm, which is given by f(S) for some non-negative monotone submodular function f. For any small constant epsilon > 0, we consider the special case that the weight of an item on every dimension is at most a (1- epsilon) fraction of the total capacity, and give a polynomial-time deterministic O(k / epsilon^2)-competitive algorithm for the problem, where k is the (column) sparsity of the weight vectors. We also show several (almost) tight hardness results even when the algorithm is computationally unbounded. We first show that under the epsilon-slack assumption, no deterministic algorithm can obtain any o(k) competitive ratio, and no randomized algorithm can obtain any o(k / log k) competitive ratio. We then show that for the general case (when epsilon = 0), no randomized algorithm can obtain any o(k) competitive ratio. In contrast to the (1+delta) competitive ratio achieved in Kesselheim et al. [STOC 2014] for the problem with random arrival order of items and under large capacity assumption, we show that in the arbitrary arrival order case, even when |w_u|_infinity is arbitrarily small for all items u, it is impossible to achieve any o(log k / log log k) competitive ratio.

We study the online submodular maximization problem with free disposal under a matroid constraint. Elements from some ground set arrive one by one in rounds, and the algorithm maintains a feasible set that is independent in the underlying matroid. In each round when a new element arrives, the algorithm may accept the new element into its feasible set and possibly remove elements from it, provided that the resulting set is still independent. The goal is to maximize the value of the final feasible set under some monotone submodular function, to which the algorithm has oracle access. For k-uniform matroids, we give a deterministic algorithm with competitive ratio at least 0.2959, and the ratio approaches as k approaches infinity, improving the previous best ratio of 0.25 by Chakrabarti and Kale (IPCO 2014), Buchbinder et al. (SODA 2015) and Chekuri et al. (ICALP 2015). We also show that our algorithm is optimal among a class of deterministic monotone algorithms that accept a new arriving element only if the objective is strictly increased. Further, we prove that no deterministic monotone algorithm can be strictly better than 0.25-competitive even for partition matroids, the most modest generalization of k-uniform matroids, matching the competitive ratio by Chakrabarti and Kale (IPCO 2014) and Chekuri et al. (ICALP 2015). Interestingly, we show that randomized algorithms are strictly more powerful by giving a (non-monotone) randomized algorithm for partition matroids with ratio Finally, our techniques can be extended to a more general problem that generalizes both the online sub- modular maximization problem and the online bipartite matching problem with free disposal. Using the techniques developed in this paper, we give constant- competitive algorithms for the submodular online bipartite matching problem.

We achieve a (randomized) polynomial-time approximation scheme (PTAS) for the Steiner forest problem in doubling metrics. Before our work, a PTAS was given only for the Euclidean plane in [G. Borradaile, P. N. Klein, and C. Mathieu, in FOCS, IEEE Computer Society, 2008, pp. 115--124]. Our PTAS also shares similarities with the dynamic programming for sparse instances used in [Y. Bartal, L. Gottlieb, and R. Krauthgamer, in STOC, ACM, 2012, pp. 663--672] and [T-H. H. Chan and S.-H. Jiang, in SODA, SIAM, 2016, pp. 754--765]. However, extending previous approaches requires overcoming several nontrivial hurdles, and we make the following technical contributions. (1) We prove a technical lemma showing that Steiner points have to be “near” the terminals in an optimal Steiner tree. This enables us to define a heuristic to estimate the local behavior of the optimal solution, even though the Steiner points are unknown in advance. This lemma also generalizes previous results in the Euclidean plane and may be of independent interest for related problems involving Steiner points. (2) We develop a novel algorithmic technique known as “adaptive cells” to overcome the difficulty of keeping track of multiple components in a solution. Our idea is based on but significantly different from the previously proposed “uniform cells” in [G. Borradaile, P. N. Klein, and C. Mathieu, in FOCS, IEEE Computer Society, 2008, pp. 115--124], where techniques cannot be readily applied to doubling metrics.

A coflow is a collection of related parallel flows that occur typically between two stages of a multi-stage compute task in a network, such as shuffle flows in MapReduce. The coflow abstraction allows applications to convey their semantics to the network so that application-level requirements (e.g., minimizing the completion time of the slowest flow) can be better satisfied. In this paper, we study the routing and scheduling of multiple coflows to minimize the average coflow completion time (CCT). We first propose a rounding-based randomized approximation algorithm, called OneCoflow, for single coflow routing and scheduling. The multiple coflow problem is more challenging as coexisting coflows will compete for the same network resources such as link bandwidths. To minimize the average CCT, we derive an online multiple coflow routing and scheduling algorithm, called OMCoflow, and prove that it has a reasonably good competitive ratio. To the best of our knowledge, this is the first online algorithm with theoretical performance guarantees which considers routing and scheduling simultaneously for multi-coflows. Compared with existing methods, OMCoflow runs more efficiently, and it avoids the problem of frequently rerouting the flows. Extensive simulations on a Facebook data trace show that OMCoflow outperforms the state-of-the-art heuristic schemes significantly (e.g., reducing the average CCT by up to 41.8% and the execution time by up to 99.2% against RAPIER [28]).

We consider the Traveling Salesman Problem with Neighborhoods (TSPN) in doubling metrics. The goal is to find a shortest tour that visits each of a given collection of subsets (regions or neighborhoods) in the underlying metric space. We give a randomized polynomial time approximation scheme (PTAS) when the regions are fat weakly disjoint. This notion of regions was first defined when a QPTAS was given for the problem in [SODA 2010: Chan and Elbassioni]. We combine the techniques in the previous work, together with the recent PTAS for TSP [STOC 2012: Bartal, Gottlieb and Krauthgamer] to achieve a PTAS for TSPN. Moreover, more refined procedures are used to improve the dependence of the running time on the doubling dimension k from the previous exp[O(1)k2] (even for just TSP) to exp[O(1)O(k log k)].

We consider the general (J, K)-secretary problem, where n totally ordered items arrive in a random order. An algorithm observes the relative merits of arriving items and is allowed to make J selections. The objective is to maximize the expected number of items selected among the K best items. Buchbinder, Jain and Singh proposed a finite linear program (LP) that completely characterizes the problem, but it is difficult to analyze the asymptotic behavior of its optimal solution as n tends to infinity. Instead, we prove a formal connection between the finite model and an infinite model, where there are a countably infinite number of items, each of which has arrival time drawn independently and uniformly from [0, 1]. The finite LP extends to a continuous LP, whose complementary slackness conditions reveal an optimal algorithm which involves JK thresholds that play a similar role as the -threshold in the optimal classical secretary algorithm. In particular, for the case K = 1, the J optimal thresholds have a nice “rational description”. Our continuous LP analysis gives a very clear perspective on the problem, and the new insights inspire us to solve two related problems. 1. We settle the open problem whether algorithms based only on relative merits can achieve optimal ratio for matroid secretary problems. We show that, for online 2-item auction with random arriving bids (the K-uniform matroid problem with K = 2), an algorithm making decisions based only on relative merits cannot achieve the optimal ratio. This is in contrast with the folklore that, for online 1-item auction, no algorithm can have performance ratio strictly larger than 1/e, which is achievable by an algorithm that considers only relative merits. 2. We give a general transformation technique that takes any monotone algorithm (such as threshold algorithms) for the (K, K)-secretary problem, and constructs an algorithm for online bipartite K-matching with random arrival order that has at least the same performance guarantee.