A quadratic lower bound for Rocchio’s similarity-based relevance feedback algorithm with a fixed query updating factor

Rocchio’s similarity-based relevance feedback algorithm, one of the most important query reformation methods in information retrieval, is essentially an adaptive supervised learning algorithm from examples. In practice, Rocchio’s algorithm often uses a fixed query updating factor. When this is the case, we strengthen the linear Ω(n) lower bound obtained by Chen and Zhu (Inf. Retr. 5:61–86, 2002) and prove that Rocchio’s algorithm makes Ω(k(n−k)) mistakes in searching for a collection of documents represented by a monotone disjunction of k relevant features over the n-dimensional binary vector space {0,1} ⁿ , when the inner product similarity measure is used. A quadratic lower bound is obtained when k is linearly proportional to n. We also prove an O(k(n−k)³) upper bound for Rocchio’s algorithm with the inner product similarity measure in searching for such a collection of documents with a constant query updating factor and a zero classification threshold.     

Finding longest increasing and common subsequences in streaming data

We present algorithms and lower bounds for the Longest Increasing Subsequence (LIS) and Longest Common Subsequence (LCS) problems in the data-streaming model. To decide if the LIS of a given stream of elements drawn from an alphabet αbet has length at least k, we discuss a one-pass algorithm using O(k log αbetsize) space, with update time either O(log k) or O(log log αbetsize); for αbetsize = O(1), we can achieve O(log k) space and constant-time updates. We also prove a lower bound of Ω(k) on the space requirement for this problem for general alphabets αbet, even when the input stream is a permutation of αbet. For finding the actual LIS, we give a ⌈log (1 + 1/ɛ)-pass algorithm using O(k^1+ɛlog αbetsize) space, for any ɛ > 0. For LCS, there is a trivial Θ(1)-approximate O(log n)-space streaming algorithm when αbetsize = O(1). For general alphabets αbet, the problem is much harder. We prove several lower bounds on the LCS problem, of which the strongest is the following: it is necessary to use Ω(n/ρ²) space to approximate the LCS of two n-element streams to within a factor of ρ, even if the streams are permutations of each other. A preliminary version of this paper appears in the Proceedings of the 11th International Computing and Combinatorics Conference (COCOON'05), August 2005, pp. 263–272.     

On the readability of monotone Boolean formulae

Golumbic et al. (Discrete Appl. Math. 154:1465–1477, 2006) defined the readability of a monotone Boolean function f to be the minimum integer k such that there exists an ∧−∨-formula equivalent to f in which each variable appears at most k times. They asked whether there exists a polynomial-time algorithm, which given a monotone Boolean function f, in CNF or DNF form, checks whether f is a read-k function, for a fixed k. In this paper, we partially answer this question already for k=2 by showing that it is NP-hard to decide if a given monotone formula represents a read-twice function. It follows also from our reduction that it is NP-hard to approximate the readability of a given monotone Boolean function f:{0,1} ⁿ →{0,1} within a factor of O(n)\mathcal{O}(n) . We also give tight sublinear upper bounds on the readability of a monotone Boolean function given in CNF (or DNF) form, parameterized by the number of terms in the CNF and the maximum size in each term, or more generally the maximum number of variables in the intersection of any constant number of terms. When the variables of the DNF can be ordered so that each term consists of a set of consecutive variables, we give much tighter logarithmic bounds on the readability.     

Tree edge decomposition with an application to minimum ultrametric tree approximation

A k-decomposition of a tree is a process in which the tree is recursively partitioned into k edge-disjoint subtrees until each subtree contains only one edge. We investigated the problem how many levels it is sufficient to decompose the edges of a tree. In this paper, we show that any n-edge tree can be 2-decomposed (and 3-decomposed) within at most ⌈1.44 log n⌉ (and ⌈log n⌉ respectively) levels. Extreme trees are given to show that the bounds are asymptotically tight. Based on the result, we designed an improved approximation algorithm for the minimum ultrametric tree.     

Partitioning a weighted partial order

The problem of partitioning a partially ordered set into a minimum number of chains is a well-known problem. In this paper we study a generalization of this problem, where we not only assume that the chains have bounded size, but also that a weight w _i is given for each element i in the partial order such that w _i ≤w _j if i ≺ j. The problem is then to partition the partial order into a minimum-weight set of chains of bounded size, where the weight of a chain equals the weight of the heaviest element in the chain. We prove that this problem is -hard, and we propose and analyze lower bounds for this problem. Based on these lower bounds, we exhibit a 2-approximation algorithm, and show that it is tight. We report computational results for a number of real-world and randomly generated problem instances.     

On the computational hardness based on linear FPT-reductions

The notion of linear fpt-reductions has been recently introduced to derive strong computational lower bounds for well-known NP-hard problems. In this paper, we formally investigate the notion of W[t]-hardness under the linear fpt-reduction, and study the structural properties of the corresponding complexity classes. Additional complexity lower bounds on important computational problems are established. Some observations on structural properties of the standard parameterized hierarchy, the W -hierarchy, are also presented. In this paper, we always assume that complexity functions are “nice” with both domain and range being non-negative integers and the values of the functions and their inverses can be easily computed.     

Fast searching games on graphs

Given a graph, suppose that intruders hide on vertices or along edges of the graph. The fast searching problem is to find the minimum number of searchers required to capture all the intruders satisfying the constraint that every edge is traversed exactly once and searchers are not allowed to jump. In this paper, we prove lower bounds on the fast search number. We present a linear time algorithm to compute the fast search number of Halin graphs and their extensions. We present a quadratic time algorithm to compute the fast search number of cubic graphs.     

Improving an upper bound on the size of k-regular induced subgraphs

This paper considers the NP-hard graph problem of determining a maximum cardinality subset of vertices inducing a k-regular subgraph. For any graph G, this maximum will be denoted by α _k (G). From a well known Motzkin-Straus result, a relationship is deduced between α _k (G) and the independence number α(G). Next, it is proved that the upper bounds υ _k (G) introduced in Cardoso et al. (J. Comb. Optim., 14, 455–463, 2007) can easily be computed from υ ₀(G), for any positive integer k. This relationship also allows one to present an alternative proof of the Hoffman bound extension introduced in the above paper. The paper continues with the introduction of a new upper bound on α _k (G) improving υ _k (G). Due to the difficulty of computing this improved bound, two methods are provided for approximating it. Finally, some computational experiments which were performed to compare all bounds studied are reported.     

The total {k}-domatic number of a graph

For a positive integer k, a total {k}-dominating function of a graph G is a function f from the vertex set V(G) to the set {0,1,2,…,k} such that for any vertex v∈V(G), the condition ∑ _u∈N(v) f(u)≥k is fulfilled, where N(v) is the open neighborhood of v. A set {f ₁,f ₂,…,f _d } of total {k}-dominating functions on G with the property that ?_i=1^df_i(v) ￡ k\sum_{i=1}^{d}f_{i}(v)\le k for each v∈V(G), is called a total {k}-dominating family (of functions) on G. The maximum number of functions in a total {k}-dominating family on G is the total {k}-domatic number of G, denoted by d_t^{k}(G)d_{t}^{\{k\}}(G). Note that d_t^{1}(G)d_{t}^{\{1\}}(G) is the classic total domatic number d _t (G). In this paper we initiate the study of the total {k}-domatic number in graphs and we present some bounds for d_t^{k}(G)d_{t}^{\{k\}}(G). Many of the known bounds of d _t (G) are immediate consequences of our results.     

On-line bin packing with restricted repacking

Semi-on-line algorithms for the bin-packing problem allow, in contrast to pure on-line algorithms, the use of certain types of additional operations for each step. Examples include repacking, reordering or lookahead before packing the items. Here we define and analyze a semi-on-line algorithm where for each step at most k items can be repacked, for some positive integer k. We prove that the upper bound for the asymptotic competitive ratio of the algorithm is a decreasing function of k, which tends to 3/2 as k goes to infinity. We also establish lower bounds for this ratio and show that the gap between upper and lower bounds is relatively small.     

Coverage with k-transmitters in the presence of obstacles

For a fixed integer k≥0, a k-transmitter is an omnidirectional wireless transmitter with an infinite broadcast range that is able to penetrate up to k “walls”, represented as line segments in the plane. We develop lower and upper bounds for the number of k-transmitters that are necessary and sufficient to cover a given collection of line segments, polygonal chains and polygons.     

Hardness of <Emphasis Type="Italic">k</Emphasis>-Vertex-Connected Subgraph Augmentation Problem

Given a k-connected graph G=(V,E) and V ^′⊂V, k-Vertex-Connected Subgraph Augmentation Problem (k-VCSAP) is to find S⊂V∖V ^′ with minimum cardinality such that the subgraph induced by V ^′∪S is k-connected. In this paper, we study the hardness of k-VCSAP in undirect graphs. We first prove k-VCSAP is APX-hard. Then, we improve the lower bound in two ways by relying on different assumptions. That is, we prove no algorithm for k-VCSAP has a PR better than O(log (log n)) unless P=NP and O(log n) unless NP⊆DTIME(n ^{O(log log n)}), where n is the size of an input graph.     

On broadcasting in unicyclic graphs

Broadcasting is an information dissemination problem in a connected network, in which one node, called the originator, disseminates a message to all other nodes by placing a series of calls along the communication lines of the network. Finding the broadcast time of a vertex in an arbitrary graph is NP-complete. The polynomial time solvability is shown only for trees. In this paper we present a linear algorithm that determines the broadcast time of any originator in an arbitrary unicyclic graph. As a byproduct, we find a broadcast center of the unicyclic graph. We also present an O(|V|+k ²) algorithm to find the broadcast time of an arbitrary unicyclic graph, where k is the length of the cycle. In the last section we give tight lower and upper bounds on broadcast time of a spanning tree based on the broadcast time of the unicyclic graph. The results of Sects. 2, 3 and most of the proofs in Sects. 2, 3 of this paper are presented by Harutyunyan and Maraachlian (Proceedings of 13th annual COCOON, pp. 372–383, 2007). All results in Sects. 4, 5 and the complete proof of Theorem 3 are new results.     

Semi-online scheduling with “end of sequence” information

We study a variant of classical scheduling, which is called scheduling with “end of sequence” information. It is known in advance that the last job has the longest processing time. Moreover, the last job is marked, and thus it is known for every new job whether it is the final job of the sequence. We explore this model on two uniformly related machines, that is, two machines with possibly different speeds. Two objectives are considered, maximizing the minimum completion time and minimizing the maximum completion time (makespan). Let s be the speed ratio between the two machines, we consider the competitive ratios which are possible to achieve for the two problems as functions of s. We present algorithms for different values of s and lower bounds on the competitive ratio. The proposed algorithms are best possible for a wide range of values of s. For the overall competitive ratio, we show tight bounds of ϕ + 1 ≈ 2.618 for the first problem, and upper and lower bounds of 1.5 and 1.46557 for the second problem. The authors would like to dedicate this paper to the memory of our colleague and friend Yong He who passed away in August 2005 after struggling with illness. D. Ye: Research was supported in part by NSFC (10601048).     

Efficient point coverage in wireless sensor networks

We study minimum-cost sensor placement on a bounded 3D sensing field, R, which comprises a number of discrete points that may or may not be grid points. Suppose we have ℓ types of sensors available with different sensing ranges and different costs. We want to find, given an integer σ ≥ 1, a selection of sensors and a subset of points to place these sensors such that every point in R is covered by at least σ sensors and the total cost of the sensors is minimum. This problem is known to be NP-hard. Let k_i denote the maximum number of points that can be covered by a sensor of the ith type. We present in this paper a polynomial-time approximation algorithm for this problem with a proven approximation ratio . In applications where the distance of any two points has a fixed positive lower bound, each k_i is a constant, and so we have a polynomial-time approximation algorithms with a constant guarantee. While γ may be large, we note that it is only a worst-case upper bound. In practice the actual approximation ratio is small, even on randomly generated points that do not have a fixed positive minimum distance between them. We provide a number of numerical results for comparing approximation solutions and optimal solutions, and show that the actual approximation ratios in these examples are all less than 3, even though γ is substantially larger. This research was supported in part by NSF under grant CCF-04080261 and by NSF of China under grant 60273062.     

Inapproximability and approximability of maximal tree routing and coloring

Let G be a undirected connected graph. Given g groups each being a subset of V(G) and a number of colors, we consider how to find a subgroup of subsets such that there exists a tree interconnecting all vertices in each subset and all trees can be colored properly with given colors (no two trees sharing a common edge receive the same color); the objective is to maximize the number of subsets in the subgroup. This problem arises from the application of multicast communication in all optical networks. In this paper, we first obtain an explicit lower bound on the approximability of this problem and prove Ω(g^1−ε)-inapproximability even when G is a mesh. We then propose a simple greedy algorithm that achieves performance ratio O√|E(G)|, which matches the theoretical bounds. Supported in part by the NSF of China under Grant No. 70221001 and 60373012.     

The canadian traveller problem and its competitive analysis

From the online point of view, we study the Canadian Traveller Problem (CTP), in which the traveller knows in advance the structure of the graph and the costs of all edges. However, some edges may fail and the traveller only observes that upon reaching an adjacent vertex of the blocked edge. The goal is to find the least-cost route from the source O to the destination D, more precisely, to find an adaptive strategy minimizing the competitive ratio, which compares the performance of this strategy with that of a hypothetical offline algorithm that knows the entire topology in advance. In this paper, we present two adaptive strategies—a greedy or myopic strategy and a comparison strategy combining the greedy strategy and the reposition strategy in which the traveller backtracks to the source every time when he/she sees a failed edge. We prove tight competitive ratios of 2 ^k+1−1 and 2k+1 respectively for the two strategies, where k is the number of failed edges in the graph. Finally, we propose an explanation of why the greedy strategy and the comparison strategy are usually preferred by drivers in an urban traffic environment, based on an argument related to the length of the second-shortest path in a grid graph. We would like to acknowledge the support from NSF of China (No. 70525004, No. 70121001 and No. 60736027), and the support from K.C. Wong Education Foundation, Hong Kong.     

The k-Canadian Travelers Problem with communication

This paper studies a variation of the online k-Canadian Traveler Problem (k-CTP), in which there are multiple travelers who can communicate with each other, to share real-time blockage information of the edges. We study two different communication levels for the problem, complete communication (where all travelers can receive and send blockage information with each other) and limited communication (where only some travelers can both receive and send information while the others can only receive information). The objective is that at least one traveler finds a feasible route from the origin to the destination with as small cost as possible. We give lower bounds on the competitive ratio for both the two communication levels. Considering the urban traffic environment, we propose the Retrace-Alternating strategy and Greedy strategy for both the two communication levels, and prove that increasing the number of travelers with complete communication ability may not always improve the competitive ratio of online strategies.     

On the complexity and approximation of non-unique probe selection using d-disjunct matrix

In this paper, we studied the MINimum-d-Disjunct Submatrix (MIN-d-DS), which can be used to select the minimum number of non-unique probes for viruses identification. We prove that MIN-d-DS is NP-hard for any fixed d. Using d-disjunct matrix, we present an O(log k)-approximation algorithm where k is an upper bound on the maximum number of targets hybridized to a probe. We also present a (1+(d+1)log n)-approximation algorithm to identify at most d targets in the presence of experimental errors. Our approximation algorithms also yield a linear time complexity for the decoding algorithms. The research of T. Znati was supported in part by National Science Foundation under grant CCF-0548895.     

On a hyperplane arrangement problem and tighter analysis of an error-tolerant pooling design

In this paper, we formulate and investigate the following problem: given integers d,k and r where k>r≥1,d≥2, and a prime power q, arrange d hyperplanes on to maximize the number of r-dimensional subspaces of each of which belongs to at least one of the hyperplanes. The problem is motivated by the need to give tighter bounds for an error-tolerant pooling design based on finite vector spaces. This work is partially supported by NSF CAREER Award CCF-0347565.