Voronoi Diagrams with a Transportation Network on the Euclidean Plane

Size: px
Start display at page:

Download "Voronoi Diagrams with a Transportation Network on the Euclidean Plane"

Transcription

1 Voronoi Diagrams with a Transportation Network on the Euclidean Plane Sang Won Bae Kyung-Yong Chwa Division of Computer Science, Department of EECS, Korea Advanced Institute of Science and Technology, Daejeon, Korea {swbae,kychwa}@jupiter.kaist.ac.kr Abstract This paper investigates geometric and algorithmic properties of the Voronoi diagram with a transportation network on the Euclidean plane. With a transportation network, the distance is measured as the length of the shortest (time) path. In doing so, we introduce a needle, a generalized Voronoi site. We present an O(nm 2 + m 3 + nm log n) algorithm to compute the Voronoi diagram with a transportation network on the Euclidean plane, where n is the number of given sites and m is the complexity of the given transportation network. Moreover, in the case that the roads in a transportation network have only a constant number of directions and speeds, we propose two algorithms; one needs O(nm+m 2 +n log n) time with O(m(n+m)) space and the other O(nm log n + m 2 log m) time with O(n + m) space. Both algorithms output the diagram with linear size. Keywords: Transportation Network; Time Metric; Voronoi Diagram; Shortest Path. 1 Introduction With a transportation network, like streets in a city, subway or bus networks, or a highway over a nation, people have become capable to move faster. Finding a shortest (time) path using such transportation is very important due to the tendency of people trying to move with the shortest duration. In this situation, let us imagine a set of sites or service stations, and suppose that one should visit any of them; here, the Voronoi diagram problem arises. This paper considers geometric and algorithmic issues about the Voronoi diagram under such situation. The Voronoi diagram on the plane with a transportation network has been studied by several researchers. Abellanas et al.[1] discussed a single innite road network as a straight line lying on the Euclidean plane. Under their setting, they showed that the Voronoi diagram for n points can be computed in O(n log n) time. In addition, transportation networks on the L 1 plane have been dealt with; Abellanas et al.[2] considered isothetic and monotone networks and, more generally, Aichholzer, Aurenhammer, and Palop[3] introduced the city Voronoi diagram, the Voronoi diagram with an isothetic transportation network on the L 1 plane and presented an algorithm computing city Voronoi diagrams based on some relations among city Voronoi diagrams, straight skeletons, and abstract Voronoi diagrams. This work is supported by grant No.R from KOSEF. 1

2 The PhD dissertation of Palop[4] gives us a comprehensive survey for algorithmic problems on proximity and location under metrics induced by transportation networks on the plane. The author considered not only nearest Voronoi diagrams but also farthest Voronoi diagrams under such metrics. Also, transportation networks of wedge shape (two half lines with a common endpoint) or of a circle were taken into account. Actually, the plane with a transportation network can be regarded as a special and degenerate case of a weighted region. Mitchell and Papadimitriou[5] posed the weighted region problem and provided an interesting solution based on Snell's law of refraction. The weighted region problem is based on a partition of the plane into polygonal regions, each associated with an individual weight as the traveling speed inside the region. The authors provided an algorithm constructing a shortest path map under their assumption in O(m 8 L) time, where m is the complexity of the polygonal partition and L denotes the precision of the problem instance. All the previous work on transportation networks has assumed that every road in a transportation network has the same speed. But, as mentioned at the beginning, several kinds of transportation networks are established on our surroundings, hence transportation networks with diverse speeds need to be considered. This paper is the rst result dealing with transportation networks with multiple roads, possibly having arbitrary directions and diverse speeds, on the Euclidean plane. 1.1 The Model A transportation network consists of several roads as non-intersecting line segments on the plane, which may share a node, which is an endpoint, in common. Each road has its own speed. This structure can be generally represented as a planar straight-line graph G = (V, E) with speed v(e) on each road e E, where V is a set of nodes and E is a set of roads. In our model, one can enter into and exit from G at any point of roads or nodes. Along any road e E, one moves at xed speed v(e). Outside the network G, one moves at unit speed in any direction. If entry or exit points are restricted to nite xed points on the roads, a Voronoi diagram for additively weighted points is yielded; this problem can be computed in an optimal time.[3] We assume that v(e) > 1 for every road e E, since, if v(e) 1, the road e does not contribute any shortest paths. Given a transportation network G, the distance with the transportation network G, called d G, is measured as the length (duration) of the shortest (time) path using roads in G. In fact, d G induces a metric on R 2 as a shortest path metric. We call the metric induced on R 2 by d G the transportation metric with G. 1.2 The Results and the Organization We present two algorithms for constructing the Voronoi diagram V G (S) under the transportation metric with G for a set S of n sites. The rst algorithm runs in O(nm 2 log n + m 3 log m) time and uses O(m(n + m)) space, where m is the number of roads in G. In the second algorithm, we improve the running time to O(nm 2 + m 3 + nm log n) time. As a special case, we can reduce the complexity when the roads in the given transportation network have only a constant number of directions and speeds. In this case, we can compute the diagram in O(nm + m 2 + n log n) time with O(m(n + m)) space, or in O(nm log n + m 2 log m) time with linear space. These algorithms are presented in Section 5 and Section 6. In doing so, we introduce a generalized site, a needle. A needle generalizes a weighted line segment so that its weight is varied linearly. A Voronoi diagram for needles satisfying a certain 2

3 property, so called non-piercing, on the Euclidean plane can be computed in O(n log n) time and in O(n) space. Arguments and details about needles are presented in Section 3. And the algorithms that we present are based on several observations on transportation networks and on needles, which will be investigated in Section 4. Under transportation metrics, it is not so easy any more to plan a shortest path, even if a source and a destination are determined. In Section 7, we discuss how to plan a shortest path under the transportation metric with G. Given a set S of n sites and a transportation network G with m roads, we will show that a shortest path to a nearest site from a given point can be reported in O(log(m+n)+r) time by point location over the diagram obtained by the algorithms, where r is the complexity of the reported path. Also, a shortest path map for a xed source point under the transportation metric can be constructed in O(m 3 ) time, in general. Summary, conclusions, future works, and open issues are given in Section 8. 2 Basic Idea In this section, we describe our basic idea to solve the problem through an observation on the plane with a transportation metric. The transportation metric has some bad properties so that computing Voronoi diagrams under it is not so easy. For example, we can easily construct a cyclic bisector with Ω(m) segments between two points; a transportation network with cyclic roads and two sites, one of which is much closer to a road than the other, see Figure 1. This violates one of necessary conditions to apply the abstract Voronoi diagram, one of the most popular and useful tools to compute Voronoi diagrams (for details about abstract Voronoi diagrams, see Subsection 3.2 or Klein[6]). Thus, we should nd a different approach to solve the problem. Figure 1: A cyclic bisector. Black solid line segments are roads, two small circles are sites, and gray curved segments are the bisector between the two sites under the transportation metric. To nd such a different approach, we have to investigate the plane with a transportation network more carefully. Let us consider a simpler case, a single road transportation network. Given a single innite road on the Euclidean plane, we know how to nd a shortest path by Abellanas et al.[1] We draw the following observation from their results. Observation 1. To reach a destination as quickly as possible, in accessing a road e, 1. The entering or exiting angle with the road e should be π/2 ± α, where sin α = 1/v(e).[1] 2. If above is impossible, access the road e through the closer node of e. 3

4 Figure 2: Illustrations of t-neighborhoods when G has a single road. Observation 1 can be also shown by calculating and analyzing some equations, but it is believed that a simple observation or intuition is sufcient to show it. First, previous results tell us that, given an innitely long road, a shortest path is either a path that uses the road and enters or exits the road with angle π/2 ± α, or a direct path. This implies the rst rule of Observation 1. Assuming that we are given a road as a segment, if one cannot directly enter or exit the road with angle π/2 ± α, then he or she had better contact a node of the road; this implies the second rule of Observation 1. By Observation 1, we can nd the t-neighborhood of p under the transportation metric with G, dened as {x d G (x, p) < t}, for any point p on the plane and t 0. We shall denote it by N G (p, t). Figure 2 shows four illustrations for the boundary of N G (p, t) when G is a single road network. Note that N G (p, t) can be represented as a union of one large disk centered at p and two needleshaped regions along the road, each of which is a union of a set of disks whose radii are linearly decreased along the road. This provides an intuition to analyze the transportation metric in a different point of view. Our strategy is to consider such a needle-shaped region produced from p as an independent Voronoi site under the Euclidean metric. In Section 3, we dene such a Voronoi site, called a needle. Then, we can compute the distance with more roads by using needles on the roads, if we can compute the distance to a needle from any point under the Euclidean metric. We will show how to compute a set of needles from given sites S, which allows us to properly analyze the given transportation network G, in Section 4 and 5. 3 Needle: a Generalized Site In this section, we introduce a needle as a more generalized Voronoi site. Needles play a very important role in computing diagrams and in planning shortest paths under the transportation metric. In Subsection 3.1, we give a formal denition of needles and show how the distance from any point to a needle can be computed. In Subsection 3.2, we discuss how to compute Voronoi diagrams for needles on the Euclidean plane. 3.1 Denition of a Needle A needle is generalized from a line segment with an additive weight. A line segment with additive weight w can be viewed as a set of weighted points with weight w in common. Note that the distance to a weighted point p with weight w p from any point q is equal to d(p, q) + w p. A needle is also a weighted line segment but weights of its points are allowed to be varied. One necessary condition for a needle is that the weight on a needle should be assigned linearly. With this property, a needle is suitable to represent N G (p, t), as noted in Section 2. The following is a denition of a needle. 4

5 (a) t 1 < t < t 2 (b) t = t 2 (c) t > t 2 Figure 3: A needle and its t-neighborhoods. Denition 1. A needle p is an additively weighted line segment s(p), whose endpoints are p 1 and p 2, with weight function w p. The weight w p (x) is given linearly from t 1 to t 2, for each x s(p) and 0 t 1 t 2, such that w p (p 1 ) = t 1 and w p (p 2 ) = t 2. The needle p is represented by a 4-tuple, (p 1, p 2, t 1, t 2 ). We may use the terms p 1 (p), p 2 (p), t 1 (p), and t 2 (p), instead of p 1, p 2, t 1, and t 2, respectively. Figure 3 shows t-neighborhoods N(p, t) of a needle p under the Euclidean metric, when t 1 < t < t 2, t = t 2, and t > t 2, case by case. As t increases from 0, N(p, t) grows from a point p 1 (p) at t = t 1 (p) and reaches p 2 (p) at t = t 2 (p) along s(p). We also dene the speed of a needle p as v(p) = d(p 1 (p), p 2 (p))/(t 2 (p) t 1 (p)) and its direction as direction toward p 2 (p) from p 1 (p). Note that if v(p) 1, N(p, t) is of shape of a disk, and thus we can regard p as a weighted point p 1 (p) with weight t 1 (t). Throughout this paper, we will draw a needle as an arrow headed in its direction, as shown in Figure 3. We now discuss the distance from any point to a needle under the Euclidean metric. For any point x R 2 and any needle p, the distance is represented as follows: d(x, p) = min y s(p) {d(x, y) + w p(y)}, where w p (y) is the weight assigned to y. w p (y) can be computed by the denition of a needle as follows: w p (y) = d(p 1 (p), y) t 1 (p) + (t 2 (p) t 1 (p)) d(p 1 (p), p 2 (p)) = t 1 (p) + d(p 1 (p), y)/v(p), for all y s(p). We dene some terms associated with a needle p. Without loss of generality, we assume that p is horizontal and p 1 is to the left of p 2. Let l 1 denote the line whose slope is tan α and which meets p 1, and l 2 denote the line whose slope is tan(π/2 + α) and which meets p 2, where α = sin 1 (1/v(p)). We dene s + (p) to be a line segment associated with p whose endpoints are p 1 and the intersection point between l 1 and l 2. Symmetrically, we dene s (p); see Figure 4. Next, we let p + be the set of points above s(p) such that, for any point x p +, a line, which is perpendicular to s (p) and meets x, intersects s (p). In this manner, we also dene p, symmetrically. Note that p + and p decompose R 2 into four disjoint and connected regions as shown in Figure 4. Lemma 1. Let p be a needle represented by (p 1, p 2, t 1, t 2 ). Then, the distance from any point x R 2 to p is as follow. d(x, s (p)) + t 1 if x p + d(x, p) = d(x, s + (p)) + t 1 if x p min{d(x, p 1 ) + t 1, d(x, p 2 ) + t 2 } otherwise 5

6 p + l 2 l 1 α s + (p) s(p) s (p) p α Figure 4: Plane decomposition around a needle p Proof. The t-neighborhood by the above equation is represented by N(s (p), t t 1 ) p + N(s + (p), t t 1 ) p (N(p 1, t t 1 ) N(p 2, t t 2 )) (R 2 \ (p + p )). One can easily show this is exactly the same as the t-neighborhood of a needle (p 1, p 2, t 1, t 2 ). By the above lemma, we can easily compute the distance from any point to a needle in constant time. 3.2 Voronoi Diagrams for Needles under the Euclidean Metric In this subsection, we compute the Voronoi diagram for a set of needles. In order to do so, we apply the abstract Voronoi diagram introduced by Klein.[6] In this model, a system (S, {J(p, q) p, q S, p q}) of bisecting curves for S is given, which is called admissible if the following conditions are fullled: 1. J(p, q) is homeomorphic to a line or empty, 2. the intersection of any two bisectors consists of nitely many components, 3. R(p, q) R(q, r) R(p, r), and 4. for any subset S S and p S, R(p, S ) is path-connected if it is nonempty, where R(p, q) = {x R 2 d(x, p) < d(x, q)}, R(p, S) = q S,p q R(p, q) and J(p, q) is the bisecting region between p and q. Needles, however, violate the rst and the fourth conditions, since needles generalize weighted line segments. Note that the bisector between two needles p and q is dened as J(p, q) = {x d(x, p) = d(x, q)}, which is computable in constant time by Lemma 1. We suggest a condition restricting needles so that the bisector system is admissible. We consider the following four cases for each needle p with respect to another needle q. 1. R(p, q) s(q) =. 2. R(p, q) s(q) and R(p, q) s(q) includes either p 1 (q) or p 2 (q), not both. 3. R(p, q) s(q) and R(p, q) s(q) includes both p 1 (q) and p 2 (q). 6

7 4. R(p, q) s(q) and R(p, q) s(q) includes neither p 1 (q) nor p 2 (q). We call two needles non-piercing if and only if, for each with respect to the other, the fourth case does not occur. Otherwise, we call them piercing. With only non-piercing needles, we are sufcient to solve our problem. We thus focus on a set of pairwise non-piercing needles. Note that the nonpiercing condition for needles generalizes the disjointness for line segments. Theorem 1. Let S be a set of pairwise non-piercing needles. Then, the system of bisecting curves for S, (S, {J(p, q) p, q S, p q}) is admissible. Proof. First, we show that J(p, q) is homeomorphic to a line. We can divide R 2 into p +, p and R 2 \ (p + p ), depending on the structures of shortest paths to p. These are four disjoint connected regions, we call them R p,1,, R p,4, see Figure 4. We can assign a function d p,i to each region R p,i such that d p,i (x) = d(x, p) for any point x R p,i and d p,i can further be interpreted as the distance to a certain weighted line segment or a certain point, which is placed out of the region R p,i. If x is on the boundary of R p,i and R p,j, d p,i (x) = d p,j (x), surely. Now, considering another needle q, R 2 can be divided into at most 16 regions {R p,i R q,j } 1 i,j 4. By the above argument, in each region we may have a bisecting curve, whose type may be a line segment, a parabolic arc or a hyperbolic arc. Further, such bisecting curves may share their endpoints only on the boundaries between regions. Hence, J(p, q) is a set of sequences of several quadratic curve segments. If p and q are non-piercing, the three regions, J(p, q), R(p, q), and R(q, p) are connected regions and have nonempty interiors, since a needle can be interpreted as a set of linearly weighted points, as described in Subsection 3.1. This implies J(p, q) is connected and acyclic, that is, homeomorphic to a line. Also, the above discussion implies nite connected intersections, since J(p, q) can be decomposed into at most 16 quadratic curve segment, each of which can be empty. For the second condition, if x R(p, q) R(q, r), then d(x, p) < d(x, q) < d(x, r), which implies that x R(p, r). Hence, R(p, q) R(q, r) R(p, r). Finally, we pick any subset S S. Since S is pairwise non-piercing, S is also pairwise nonpiercing. Thus, for any p S, R(p, S ) s(p), say U p, is connected or empty, and every p s(p) \ U p is dominated by S \ {p}. Hence, R(p, S ) is connected, further, path-connected, if it is nonempty. Constructing abstract Voronoi diagrams has been considered by several researchers, Klein[6], Mehlhorn et al.[7], Klein et al.[8], and Dehne and Klein [9]. All presented algorithms take an optimal time and space. Thus, we conclude the following corollary as a result of this section. Corollary 1. The Voronoi diagram for a set of n pairwise non-piercing needles on the Euclidean plane can be computed in O(n log n) time and O(n) space. In fact, Aichholzer, Aurenhammer, and Palop[3] introduced the concept similar to non-piercing needles. The authors discussed their concepts in the wavefront model and they dened a needle as a gure produced by the wavefront of a site while touching road segments. Note that, in the wavefront model, each site p sends out a wavefront and the interference pattern of these wavefronts contributes the diagram. In contrast with their approach, we dene a needle as an independent Voronoi site. 4 Relation between Transportation Networks and Needle Sets In this section, we discover a relation between a given transportation network and a set of needles. More precisely, we will show that there exists a set S of needles such that its Voronoi diagram V(S) 7

8 under the Euclidean metric induces the diagram V G (S) for a set S of sites. Further, we will show that we can make the set S pairwise non-piercing through some procedures, in Subsection 4.1. Also, several lemmas that reveal computational and combinatorial properties about transportation networks and needles will be presented in Subsection 4.2. Discussions and results in this section will be helpful to develop algorithms in the next section. 4.1 The Reducibility A point p produces at most two new needles on a road e (see Figure 2). These needles are produced either at a node of e or at an intersection point of the road with a line whose angle with e is π/2 ± α and which meets p by Observation 1. One important fact is that the produced needles are also Voronoi sites and thus they possibly produce new needles on other roads. We will call such a newly produced needle a needle produced on a road from a (parent) needle. Note that a needle also produces in general at most two new needles on a road, as a point does. Suppose that a needle produces three needles on a road. Then, two of the three should be in the same direction and one of these two needles must be dominated by the original needle and/or other newly produced ones, since the speeds of the three needles are the same and their neighborhoods are convex under the Euclidean metric (see Figure 2 and 3). More formally, we dene σ G (p) to be a set of needles produced from a needle p on all roads in G. For a set X of needles, σ G (X) = p X σ G(p). In order to obtain all needles produced from a site p, we will apply σ G ( ), repeatedly. We let σg k (p) = σ G(σ k 1 G (p)) and σ0 G (p) = {(p, p, 0, 0)}. We then let Sp k denote k i=0 σi G (p) and S p denote Sp. Now, we are ready to prove the following theorem. Theorem 2. Given a transportation network G and a set of sites S, there exists a set S of needles such that V(S) induces V G (S), i.e., for any p S, there exists a site q S such that R(p, S) R G (q, S). Proof. First, we dene d k G (p, q) be the length of a shortest path from p to q where the path passes through at most k roads. Surely, d G (p, q) = d G (p, q). We claim that d k G (p, q) = d(p, Sk q ), which directly implies the theorem. There exists a needle q Sq k such that d(p, q) = d(p, Sq k ) and q σg l (q), for any point p and 0 l k. Let us consider a sequence of needles, q l = q, q l 1,, q 1, such that q i is produced from q i 1 and q i σg i (q). With these needles, we construct a path P from p to q passing through l roads. P starts from p and goes to the points y l on s(q l ) which obeys the rule of Observation 1. Next, P passes p 1 (q l ) along the road, on which q l is produced (see Figure 5(a)). Recursively, P passes y i and p 1 (q i ) next to p 1 (q i+1 ) and nally reaches q next to p 1 (q 1 ). Since each q i is produced on a road, P passes through l roads, which is at most k. Hence, d k G (p, q) d(p, Sk q ). p y l q l q l 1 y l 1 v i q i y i x i q i 1 v i 1 y i 1 x i 1 (a) (b) Figure 5: Proof of Theorem 2 8

9 Let us consider a path P of length d k G (p, q) from p to q and assume that P passes through l roads, for 0 l k. We then construct l needles q 1,, q l, inductively, such that q i σg i (q) and d(p, q l ) = d k G (p, q). P consists of l segments of roads e l, e l 1,, e 1 in order, each of which starts at y i and ends at x i on road e i. Also, v i denotes a node of e i in direction toward y i from x i (see Figure 5(b)). We construct q i as follows: q 1 = (x 1, v 1, d(x 1, q), d(x 1, q)+ 1 v(e 1 ) d(x 1, v 1 )) and q i = (x i, v i, d(x i, q i 1 ), d(x i, q i 1 )+ 1 v(e i ) d(x i, v i )). For convenience, let q 0 be (q, q, 0, 0). Then, q i σg i (q) since P is a shortest path that uses at most k roads and it obviously obeys the rule of Observation 1, locally. Surely, d(x 1, q 0 ) = d 0 G (x 1, q) = d(x 1, q). If d(x i, q i 1 ) = d i G (x i, q), a sub-path P of P from x i to q is a shortest path passing through i roads. We construct a path P of length d(x i+1, q i ) from x i+1 to q. P is obtained by adding the path x i+1 y i x i at the beginning of P. Then, d(x i+1, q i ) = d i+1 G (x i+1, q), since P passes through i + 1 roads and a sub-path of P. By induction, d(p, q l ) = d l G (p, q) = dk G (p, q), implying that dk G (p, q) d(p, Sk q ). Finally, we show that d k G (p, q) = d(p, Sk q ), and thus d G (p, q) = d(p, S q ). With taking S = p S S p, this claim implies that, for any p S, there exists a site q S such that R(p, S) R G (q, S). Theorem 2 shows the existence of a set of needles noted at the end of Section 2. But, S may not be pairwise non-piercing and further have innitely many needles by the construction from the proof of Theorem 2. Thus, we should reduce the number of needles and make them pairwise non-piercing in order to efciently compute V(S). For any p S, we call p effective if R(p, S), otherwise we call p ineffective. Lemma 2. There exists a set S of needles, which are effective and pairwise non-piercing, such that V(S ) = V(S). Proof. First, we remove all ineffective needles from S. We let S e S be the set of all effective needles in S. Trivially, V(S e ) = V(S). We will construct S from S e by making it pairwise nonpiercing without any change in its diagram. We can partition S e into S 0, S 1,, S m such that S 0 = {(p, p, 0, 0) p S} and all needles in S i lie on e i, for 1 i m, if all the roads in G are given as {e 1,, e m }. For each needle p S i, 1 i m, we can nd (at most) one needle q S i, the closest needle in the direction of p from p 1 (p). p stretches to a node of e i but the part over p 1 (q) does not contribute V(S e ), since both p and q are effective and their speeds are the same as v(e i ). We thus cut p by setting p 2 (p) to be p 1 (q) and t 2 (p) to be t 1 (p) + 1 v(e i ) d(p 1(p), p 1 (q)). We denote the set of these new needles from S i by Si and S 0 S1 S m by S. Since we cut only a useless part of every needle from S e, V(S ) = V(S e ). Now, we show that S is pairwise non-piercing. S 0 is clearly pairwise non-piercing, since it is a set of points. Si is also pairwise non-piercing, since all needles in S i lie on a line including a road e i and they are disjoint. Let us consider two needles in S, p and q, such that they are produced on two distinct roads. Note that one of p and q may be a point in S. We assume that p pierces q and q is produced on a road e i. First, q is not a needle produced from p, since p and q should be non-piercing if q is a needle produced from p. If a needle p produced on e i from p is effective, q cannot dominate any point on e i over p 1 (p ), and thus q is not pierced. Hence, no needles on e i from p are effective. This implies that q dominates two possible starting points of needles produced from p, and thus q dominates all points between the two points. Consequently, p does not pierce q, a contradiction. Therefore, S is pairwise non-piercing. 9

10 Theorem 2 and Lemma 2 say that the Voronoi diagram with a transportation network can be represented by that for an appropriate set of pairwise non-piercing needles under the Euclidean metric. Also, Lemma 2 provides a method to make a set of needles on roads to be pairwise non-piercing. More computational and quantitative issues about needles produced on the given network G are considered in the next subsection. Further, these theorem and lemma imply that a shortest path with a transportation network can be represented as a sequence of needles. More on shortest path planning under the transportation metric will be considered in Section Needles Produced on a Road from a Parent Needle We already know how a point produces at most two needles on a road by Observation 1. This subsection generalizes such an argument and we thus focus on how a needle produce needles on a road and on how many needles we have to handle to solve the problem. Lemma 3. If a needle p does not intersect a road e, needles produced on e from p are at most two of the following six needles; four needles produced from weighted points p 1 (p) with weight t 1 (p) or p 2 (p) with weight t 2 (p), and two needles starting from a node of e. Proof. Since a needle can be viewed as a set of weighted points, we consider all needles from such weighted points. And we assume e is horizontal without loss of generality. First, let S l s(p) be a set of points such that, for all x S l, there exists a needle q produced on e from x such that its direction is to the left. For any x S l, let q be a needle produced on e from x with weight w p (x) such that its direction is to the left. Then, we have t 1 (q) = w p (x) + d(x, p 1 (q)) and we minimize the time reaching the left node p of e, t 1 (q) + 1 v(e) d(p 1(q), p), over all x S l, where v(e) is the speed of e. x 2 S l x p x 1 x 1 p 1 (q) p 1 (q 1 ) e x h l α Figure 6: Minimizing the length to the left endpoint of e. Note that S l is a line segment whose endpoints are x 1 and x 2. Any x S l can be represented by #» x = x #» 1 + t #» d, for 0 t 1, where #» x is a vector with the same coordinates as x and #» d = x2 #» x #» 1. Then, we can say that w p (x) = w p (x 1 ) + w, d(x, p 1 (q)) = d(x, p 1 (q 1 )) + l and d(p 1 (q), p) = d(p 1 (q 1 ), p) + h, as in Figure 6. Furthermore, w, l and h are of the form ct, where c is a constant, 1 which implies that the equation is linear over 0 t 1. Hence, it is minimized when x is an endpoint of S l since p does not intersect e. Assuming that if at x = x 1 the equation is minimized, either x 1 is an endpoint of p or p 1 (q 1 ) is an endpoint of e. Consequently, the left-headed needle produced from p is a needle from an endpoint of p with its weight or a needle starting from a node of e. An analogous argument is applicable to the right-headed needle. We thus show this lemma. 1 Exactly, w = t(w p(x 2) w p(x 1)), l = t #» d sin θ #» d sec α and h = t #» d (cos θ #» d sin θ #» d tan α), where is the L 2 norm and θ #» d is the angle dened by #» d. 10

11 Lemma 4. Let p be a needle produced on a road e from a needle q. For another road e E, if p does not dominate any node of e or e, no needles produced on e from p are effective. Proof. Suppose that p does not dominate any node of e or e. Then, there exists a neighboring needle in S to the growing direction of p, as shown in Figure 7. However, since p does not dominate any node of e or e, the needles produced from p 2 (p) with weight t 2 (p) and the needle starting from a node of e are ineffective. Further, the needles from p 1 (p) with weight t 1 (p) are dominated by q, since they are closer to q than to p. By Lemma 3, we cover all candidates of effective needles produced on e from p. Thus, no needles produced on e from p are effective. q e p e Figure 7: No effective needles produced on e from p. These two lemmas reduce the number of necessary applications of σ G ( ). Thus, they will be very helpful to devise an algorithm to compute S in Subsection 5.1. Further, we can show the number of needles in S by Lemma 3 and 4. Lemma 5. S = O(m(n + m)), where n = S and m = E. Moreover, this upper bound for S is tight for the worst case. Proof. For a needle p, σ G (p) is at most 2m, if p is disjoint from all the roads. If a needle q σ G (p) does not dominate any node in V, then no needles in σ G (q) are effective by Lemma 4. In other words, we consider only needles dominating at least a node as an independent site. This implies that S = O(m(n + m)). Further, the tightness of the upper bound for S can be shown by a simple example, see Figure 8. The gure depicts a transportation network, a set of sites and a set S of produced needles; m/2 roads are fanned out and the remainders are placed horizontally as the gure. If we properly adjust the speed of each road, we can obtain m 2 (n + m 2 ) + n = O(m(n + m)) number of needles in S. 5 Algorithms In this section, we consider more algorithmic issues and present simple and efcient algorithms for computing the Voronoi diagram for sites S with a given transportation network G. First, we will give 11

12 Figure 8: A tight example for the upper bound of S. Circles with empty interior are sites and arrows are needles produced on the roads. One site is placed at the top of the gure and the others are placed below. an algorithm with time complexity O(nm 2 log n + m 3 log m) in Subsection 5.1. Next, an improved algorithm that takes O(nm 2 + m 3 + nm log n) time will be presented in Subsection An Algorithm As noted earlier, since S can include possibly innitely many needles, it is inadequate to compute S from S. The effectiveness of a needle is dened with respect to all other needles in S. However, the effectiveness of a needle p can be tested only with effective needles already emitting wavefronts at time t 1 (p). Thus, our algorithm simulates the wavefront model where the wavefronts sweep the plane. Recalling the wavefront model, each site starts sending out a wavefront at time 0 and the interference pattern between such wavefronts constitutes the Voronoi diagram. While simulating the wavefront model, at time t 1 (p), we test the effectiveness of p and we update some data structures for p, only if p is effective. In this way, we compute S directly, not from S. In order to do so, the algorithm handles two kinds of events, dened as certain situations in the wavefront model; one occurs when a needle produced on a road from another effective needle starts emitting a wavefront, called a birth event, and the other occurs when a wavefront sent out from a needle reaches a node, called a node event. From each event, we can ascertain when and where it occurs, the associated site (or needle), and so on. We will call an event effective if its associated needle dominates the point where the event occurs. To handle these events, we need two types of data structures: Q is an event queue implemented as a priority queue. The priority of an event e is its occurring time. Q supports inserting, deleting, and extracting-minimum in logarithmic time with linear space. T 1, T 2,, T m are balanced binary search trees, each associated with e i, where the road set E is given as {e 1, e 2,, e m }. Each T i stores needles on e i in order. The precedence is the position of the starting point of a needle and ties are broken by their directions. T i supports inserting and deleting of a needle in logarithmic time, and also a linear scan for needles currently 12

13 in T in linear time and space. The linear scan is possible by adding some links among the elements. For details about the above data structures, refer the book by Cormen et al.[10] Also, we discuss how an event is computed or created. A birth event b on a road e from a needle p is directly associated with a needle q produced on e from p. b occurs at time t 1 (q) at p 1 (q) on the road e. And, a node event a on a node v from a needle p occurs at time d(v, p) on v Computing S Now, we are ready to describe our rst algorithm. At the beginning of the algorithm, we compute the associating birth event with each needle in σ G (p) for all p S and insert it into Q. And, for each node v V, we compute the node event for the nearest site from v, insert it into Q, and set the variable event(v) to be the node event. Once the above initialization is done, we extract the upcoming event from Q and process it as follows repeatedly while Q is not empty. If a birth event b on e i occurs, we rst test the effectiveness of its associated needle p, and then insert it to T i. The effectiveness test can be done by checking at most two neighbor needles in T i. More precisely, we nd at most two neighbors of p by the dened order in T i, and test whether two neighbors dominate p 1 (p). (Note that this test is necessary but not sufcient, i.e., some ineffective needles may pass this test. This however does not increase the asymptotic number of needles we handle. We will discuss about this argument later.) If p passes the effectiveness test, we insert p into T i and compute node events from p. Note that event(v) is the currently earliest node event on v in Q. We let a be the node event on v newly computed from p. Then, only if a will occur earlier than event(v), we delete event(v) from Q, insert a into Q and set event(v) as a. When a node event a on a node v occurs, we compute σ G (p), where p is the needle associated with a. Then, from all needles in σ G (p), we compute the associating birth events and insert them into Q. After the event processing, all effective needles are stored in each T i. We then extract them with proper cutting, as described in the proof of Lemma 2. We denote S 1 i m T i by S a and the resulting set of needles after running the algorithm by S a. The following lemma guarantees the correctness of the algorithm. Lemma 6. Let S e be a set of effective needles in S. Then, S a is pairwise non-piercing and S e S a S. Proof. While running the algorithm, all the needles inserted into T i are obtained by applying σ G ( ) from any site p S, which implies that S a S. The algorithm lters ineffective needles efciently by Lemma 4 and by the effectiveness testing in processing a birth event, which is necessary. Hence, all needles in S \ S a are ineffective and S e S a. Note that all ineffective needles are totally dominated by other effective ones. To effective needles in S a, we can apply an analogue to the proof of Lemma 4. Thus, S a is pairwise non-piercing. Finally, we compute V(S a) by using Corollary 1 such that V(S a) = V(S ) = V(S) by Lemma 6 and 2. By Theorem 2, V(S) is a super diagram of V G (S) and thus we can easily obtain V G (S) from V(S) or V(S a). Figure 9 shows a brief description of the algorithm discussed above. 13

14 Input: A set S of sites and a transportation network G. Output: The Voronoi diagram V G (S). 1. Initialize Q and T i, 1 i m. 2. Compute all birth events from σ G (p) for all p S and insert them into Q. 3. For each node v V, compute the node event from the nearest site from v, insert it into Q, and set event(v) to be the node event. 4. While Q is not empty, 5. Extract the earliest event from Q. 6. Handle the event according to its type. 7. Let S a = S 1 i m T i. 8. Make S a pairwise non-piercing by Lemma 2. Let S a be the result. 9. Compute V(S a). 10. Extract V G (S) from V(S a). Figure 9: Description of the algorithm to compute V G (S) Analysis Our algorithm depends on the number of handled events and on the number of needles in S a. In fact, the number of events is O(m(n + m)), since the algorithm applies σ G ( ) only O(n + m) times and σ G (p) = O(m) for any needle p, and all the birth events are generated by a produced needle by σ G ( ). Note that the number of handled node events is exactly the number of nodes in V. Also, S a = O(m(n + m)) by Lemma 5 and 6. Finally, we conclude our main theorem. Theorem 3. Let G be a transportation network with m roads and S be a set of n point sites. The Voronoi diagram V G (S) for S under the transportation metric with G can be computed in O(nm 2 log n + m 3 log m) time and O(m(n + m)) space. Proof. Handling an event takes O(m log(n + m)) time. Thus, computing S a takes O(m 2 (n + m) log(n + m)) time which is the bottleneck of the algorithm. For any other steps of the algorithm, O(m(n + m) log(n + m)) time is sufcient. For the space complexity, the algorithm uses only linear space on the number of handled events and produced needles, which is O(m(n + m)). 5.2 An Improved Algorithm The algorithm in the previous subsection can be improved based on further observations on transportation networks. This subsection will show a new and important property of shortest paths under the transportation metric and an improved algorithm based on the new property Primitive Paths Here, we introduce a concept of a primitive path. 14

15 Denition 2. A path P under the transportation metric is called primitive, if the following are satised: 1. P contains no nodes in V in its interior. 2. P passes through at most one road. We call a primitive path with the shortest length a shortest primitive path. Lemma 7. Given a transportation network G on the Euclidean plane, for two points p, q R 2, there exists a shortest path P from p to q with G such that P is a sequence of shortest primitive paths whose endpoints are p, q, or nodes in V. Proof. Let P be a p-q path passing through a road e and next another road e. And let us assume that P contains neither a node of e nor a node of e. Then, Lemma 4 implies that there exists a path P no longer than P such that P passes through at least a node of e or e. This consequently implies that there exists a shortest path such that, if the path goes through roads e 1,, e k in order, it passes through at least one node of e i or e i+1 between every e i and e i+1 for 1 i k 1. Such a shortest path can be partitioned into a sequence of shortest primitive paths whose endpoints are p, q, or nodes in V. By Lemma 7, we can compute a shortest path for two given points, p and q. All shortest primitive paths whose endpoints are p, q, or nodes, can be generated since the number of such paths is exactly ( V + 2) 2 = O(m 2 ) and this can be represented as an edge-weighted complete graph whose vertex set is V {p, q} Modication Recall that a bottleneck of our algorithm is the birth event processing part. We handle O(m(n + m)) number of birth events and it takes O(m log(n + m)) time to process each event. More precisely, in processing a birth event b, we compute the node event on v from the needle p b associated with b and update event(v), for every node v V. However, updates of event(v) does not occur so much frequently, O(m) times for every birth event, indeed. We let G P = (V P, E P ) be a graph with weight w P over E P such that V P = V S and E P = {pv or vp p V S and v V }. We assign a weight w P (pq) to every pq E P as the length of a shortest primitive path from p to q. We denote by e(p, q) the road used by a shortest primitive path from p to q. Observe that no road may be used by shortest primitive paths and then e(p, q) is not dened. We modify the birth event processing part. Note that all the birth events are generated at the initialization step or at processing a node event. Let us consider a birth event b on road e i such that b is generated from p which is a node or a site in V S. We also assume that b is passed the effectiveness test. Then, for any node v V, v p, if e(p, v) e i, there should be a shortest primitive path from p to v that passes through not e i but another road. Thus, we do not need to update event(v) in that case. Hence, we can reduce the number of updates of node events. The following is a brief description of a birth event processing. If a birth event b on e i occurs, we rst test the effectiveness of its associated needle p, and then insert it to T i if the test is passed. And we assume that b is generated from p which is a node or a site in V S. For every node v V, if e(p, v) = e i, we compute the node event a on v from p and only if a will occur earlier than event(v), we delete event(v) from Q, insert a into Q, and set event(v) as a. Otherwise, if e(p, v) e i, we do nothing. 15

16 The above procedure takes O(m + l log(n + m)) time, where l is the number of updated node events. Since only one node event on a node is handled, the total number of updated node events is at most O(m(n + m)). Hence, with this new event processing procedure, the time complexity of the algorithm is reduced to O(m 2 (n + m) + m(n + m) log(n + m)) = O(nm 2 + m 3 + nm log n). Only remaining issues are how to compute the weight w P and how to evaluate e(v, w) for any vw E P. First, we can obtain w P (v, w) by comparing the lengths of primitive paths from v to w over all possible roads in E. It takes O(m) time for every edge vw E P. We further obtain e(v, w) while computing w P (v, w). If O(m(n + m)) storage is allowed to store e(v, w), we can evaluate e(v, w) in constant time. In summary, with O(m 2 (n + m)) preprocessing time and O(m(n + m)) space, we can compute the graph G P with the weight w P and we can evaluate e(v, w) in constant time for every vw E P. Consequently, we have a faster algorithm to compute the Voronoi diagram V G (S) for a set S of sites with a transportation network G. Theorem 4. Let G be a transportation network with m roads and S be a set of n point sites. The Voronoi diagram V G (S) for S under the transportation metric with G can be computed in O(nm 2 + m 3 + nm log n) time and O(m(n + m)) space. 6 Restriction on Transportation Networks In the previous section, we dealt with transportation networks as planar straight-line graphs with individual speeds and algorithms for computing the Voronoi diagram with such a transportation network. In this section, we consider a special case of transportation networks, which will lead to a reduction of running time and required storage in computing Voronoi diagrams. Here, we assume that there are at most k directions and speeds for the roads. We will obtain a better bound for S than that obtained in the previous section. Let d #» 1, d #» 2,, d #» k denote the k possible directions of the roads and v i be the speed associated with the direction d #» i for each i, 1 i k. Also, we denote by E i the set of roads in E whose directions are d #» i. Note that, in any E i, all the roads are parallel and their speeds are the same as v i. p e e q r Figure 10: Parallel roads with the same speed. r is always ineffective. Lemma 8. Given transportation network G with k possible directions and speeds of the roads, S = O(k(n + m)). Proof. In order to prove this lemma, it is sufcient to show that the number of effective needles in σ G (p) that are not starting from a node is at most O(k) for any needle p. 16

17 Without loss of generality, assume that #» d i is horizontal. For any needle p and two roads, e, e E i such that they are below p and e is closer than e to p, we let q be a needle produced from p on e and r be a needle produced on e. We also assume that q and r are not starting from a node. Then, we observe that N(r, t) N(q, t) for any positive t, since the two roads have the same speed of v i ; see Figure 10. This implies that r is not effective and the number of effective needles in σ G (p) that are not starting from a node is at most 4k. By Lemma 8, we can reduce the complexity of computing V G (S) with a minor modication from the algorithm presented in the previous section. The algorithm specied in Section 5 inserts O(m) birth events for each parent needle into Q. Recall that a birth event is created either at the initializing part or in handling a node event. We modify these parts to insert only O(k) birth events for each parent needle in O(k) time and we may need some preprocessing to do so. At the preprocessing step, we compute at most 4k roads for each p S V, which are candidates on which at most 4k effective needles are produced from p. We denote the set of such roads by E(p). First, we construct four maps M i,1,, M i,4 for each direction d #» i, 1 i k. Before describing how to construct them, we dene four directions d #» i,1,, d #» i,4 obtained by rotating d #» i by angles π/2 ± α i and π/2 ± α i, where sin α i = 1/v i. Note that these four directions induce shortest paths using any road in E i by Observation 1. We are now ready to construct the map M i,j, for 1 i k and 1 j 4. From every node incident to roads in E i, we draw along a ray with direction d #» i,j and stop drawing when reaching another road in E i, see Figure 11. With the roads in E i, we then obtain a polygonal subdivision of the plane with O( E i ) size and this is computable in O( E i log E i ) time by applying the plane sweep. Thus, we can construct such 4k maps in O(m log m) time and O(m) space. In addition, each region is labeled as the road to the direction d #» i,j. Note that there is one unlabeled region that bounds the other regions. d i,j di Figure 11: The map M i,j with E i and #» d i,j With the maps M i,j, we can compute E(p) for any point p R 2 by point location over each M i,j, in total 4k times. Let e be the road obtained by point location of p over M i,j. Then, by proof of Lemma 8, any needle produced from p on any road over e in direction d #» i,j is not effective. Hence, E(p) can be computed as a set of such roads obtained by point locations of p over the 4k maps. Finally, we have E(p) for every point p S V in O(k(m + n) log m) time and O(k(n + m)) space. As mentioned before, we now produce only O(k) birth events for each site at the initializing part and for each node event in O(k) time; This is done by checking only O(k) roads in E(p) instead of all roads in E, for any site or node p. In this way, the number of total birth events is reduced to O(k(n + m)) and also the complexity of the algorithm is. In Section 5, we proposed two algorithms. Based on each algorithm, Theorem 5 gives two algorithms and the analysis of them. 17

18 Theorem 5. Given transportation network G with k possible directions and speeds of the roads, V G (S) can be computed in O(knm+km 2 +kn log n) time with O(m(n+m)) space, or in O(nmk log n+ km 2 log m) time with O(k(n + m)) space. Proof. Here, we present two different algorithms based on the algorithms in Subsection 5.1 and 5.2. Based on the algorithm in Subsection 5.1. According to above discussion, we have O(k(n + m)) number of birth events and O(m) number of node events. By proof of Theorem 3, handling an event takes O(m log(n + m)) time, which implies that S a can be computed in O(km(n + m) log(n + m)) time. Also note that S a = O(k(n + m)) by Lemma 6 and 8. Hence, we can compute V G (S) under the assumption of this theorem in O(nmk log n + km 2 log m) time with O(k(n + m)) space. Based on the algorithm in Subsection 5.2. Using the maps M i,j, we can build the graph G P in O(km(n + m)) time. We handle O(k(n + m)) birth events and O(m) node events. By proof of Theorem 4, under the assumption of this theorem, we can compute V G (S) in O(km(n + m) + (M + k(n + m)) log(n + m)) time with O(m(n + m)) space, where M is the number of updates of node events during the modied algorithm. Actually, M = O(k(n + m)), since E(p) = O(k) and thus E(p) E(q) = O(k) for any p, q R 2. Hence, we have an O(knm + km 2 + kn log n) algorithm to compute V G (S) under the assumption of this theorem. Theorem 5 says that if we regard k as a constant (e.g., an isothetic network) the diagram can be computed in O(nm + m 2 + n log n) time or in O(nm log n + m 2 log n) time. We note that this restriction on a transportation network leads to a linear size of the resulting diagram. This result is analogous to the previous result by Aichholzer, Aurenhammer, and Palop[3] They also maintained a linear size of the diagram by restricting their transportation network as an isothetic network with the same speed for the roads on the L 1 plane. 7 Shortest Path Planning 7.1 Shortest Path to the Nearest Site By point location over Voronoi diagrams, we can easily nd the nearest site from any point in the plane. Under conventional metrics, for example, the Euclidean metric, the L 1 metric, and so forth, for any point p R 2, we can compute the nearest site in O(log n) time with the Voronoi diagram when n sites are given, and the complexity of shortest paths is constant. Thus, we can nd a shortest path to the nearest site, naturally and trivially. However, under transportation metrics, it becomes more complicated. The complexity of shortest paths is no longer constant but O(m) for the worst case, where m is the complexity of the given transportation network. In this subsection, we discuss how to construct a shortest path to the nearest site using diagrams obtained by our algorithms. With the Voronoi diagram V G (S) for a set S of sites under the transportation metric with G, we cannot plan a shortest path from a query point to the nearest site, although under conventional metrics we can. This is due to the fact that a region in V G (S) has insufcient information of the structures of shortest paths to the site associated with the region. However, fortunately, we have another diagram V(S ) obtained while running the algorithm. Note that the region R(p, S ) for each p S is a set of points from which the nearest site is p when R(p, S ) R G (p, S), and shortest paths to p have the same structures; details are shown in proof of Theorem 2. To construct a shortest path efciently, we maintain parent(p) for any p S in addition; parent(p) is dened as the parent needle in S, from which p is produced. The relation between p and parent(p) makes a forest structure that contains n rooted trees whose roots indicate sites in 18

Travel Time Distances Induced by Transportation Networks and General Underlying Distances *

Travel Time Distances Induced by Transportation Networks and General Underlying Distances * JOURNAL OF INFORMATION SCIENCE AND ENGINEERING 30, 1445-1461 (2014) Travel Time Distances Induced by Transportation Networks and General Underlying Distances * SANG WON BAE 1 AND KYUNG-YONG CHWA 2 1 Department

More information

k th -order Voronoi Diagrams

k th -order Voronoi Diagrams k th -order Voronoi Diagrams References: D.-T. Lee, On k-nearest neighbor Voronoi Diagrams in the plane, IEEE Transactions on Computers, Vol. 31, No. 6, pp. 478 487, 1982. B. Chazelle and H. Edelsbrunner,

More information

Bichain graphs: geometric model and universal graphs

Bichain graphs: geometric model and universal graphs Bichain graphs: geometric model and universal graphs Robert Brignall a,1, Vadim V. Lozin b,, Juraj Stacho b, a Department of Mathematics and Statistics, The Open University, Milton Keynes MK7 6AA, United

More information

Shortest paths with negative lengths

Shortest paths with negative lengths Chapter 8 Shortest paths with negative lengths In this chapter we give a linear-space, nearly linear-time algorithm that, given a directed planar graph G with real positive and negative lengths, but no

More information

Fly Cheaply: On the Minimum Fuel Consumption Problem

Fly Cheaply: On the Minimum Fuel Consumption Problem Journal of Algorithms 41, 330 337 (2001) doi:10.1006/jagm.2001.1189, available online at http://www.idealibrary.com on Fly Cheaply: On the Minimum Fuel Consumption Problem Timothy M. Chan Department of

More information

The Great Wall of David Shin

The Great Wall of David Shin The Great Wall of David Shin Tiankai Liu 115 June 015 On 9 May 010, David Shin posed the following puzzle in a Facebook note: Problem 1. You're blindfolded, disoriented, and standing one mile from the

More information

Computing a Minimum-Dilation Spanning Tree is NP-hard

Computing a Minimum-Dilation Spanning Tree is NP-hard Computing a Minimum-Dilation Spanning Tree is NP-hard Otfried Cheong 1 Herman Haverkort 2 Mira Lee 1 October 0, 2008 Abstract In a geometric network G = (S, E), the graph distance between two vertices

More information

On Powers of some Intersection Graphs

On Powers of some Intersection Graphs On Powers of some Intersection Graphs Geir Agnarsson Abstract We first consider m-trapezoid graphs and circular m-trapezoid graphs and give new constructive proofs that both these classes are closed under

More information

k-degenerate Graphs Allan Bickle Date Western Michigan University

k-degenerate Graphs Allan Bickle Date Western Michigan University k-degenerate Graphs Western Michigan University Date Basics Denition The k-core of a graph G is the maximal induced subgraph H G such that δ (H) k. The core number of a vertex, C (v), is the largest value

More information

Computing a Minimum-Dilation Spanning Tree is NP-hard

Computing a Minimum-Dilation Spanning Tree is NP-hard Computing a Minimum-Dilation Spanning Tree is NP-hard Otfried Cheong Herman Haverkort Mira Lee Dept. of Computer Science, Korea Advanced Institute of Science & Technology, Daejeon, South Korea. Email:

More information

1 Some loose ends from last time

1 Some loose ends from last time Cornell University, Fall 2010 CS 6820: Algorithms Lecture notes: Kruskal s and Borůvka s MST algorithms September 20, 2010 1 Some loose ends from last time 1.1 A lemma concerning greedy algorithms and

More information

29 Linear Programming

29 Linear Programming 29 Linear Programming Many problems take the form of optimizing an objective, given limited resources and competing constraints If we can specify the objective as a linear function of certain variables,

More information

Computational Geometry

Computational Geometry Voronoi diagrams Voronoi diagram Given some trees, seen from above, which region will they occupy? Voronoi diagram Given some trees, seen from above, which region will they occupy? Voronoi diagram Given

More information

Fast algorithms for collision and proximity problems involving moving geometric objects

Fast algorithms for collision and proximity problems involving moving geometric objects Fast algorithms for collision and proximity problems involving moving geometric objects Prosenjit Gupta Ravi Janardan Michiel Smid August 22, 1995 Abstract Consider a set of geometric objects, such as

More information

Tree-width and planar minors

Tree-width and planar minors Tree-width and planar minors Alexander Leaf and Paul Seymour 1 Princeton University, Princeton, NJ 08544 May 22, 2012; revised March 18, 2014 1 Supported by ONR grant N00014-10-1-0680 and NSF grant DMS-0901075.

More information

Faster Algorithms for some Optimization Problems on Collinear Points

Faster Algorithms for some Optimization Problems on Collinear Points Faster Algorithms for some Optimization Problems on Collinear Points Ahmad Biniaz Prosenjit Bose Paz Carmi Anil Maheshwari J. Ian Munro Michiel Smid June 29, 2018 Abstract We propose faster algorithms

More information

(x 1, y 1 ) = (x 2, y 2 ) if and only if x 1 = x 2 and y 1 = y 2.

(x 1, y 1 ) = (x 2, y 2 ) if and only if x 1 = x 2 and y 1 = y 2. 1. Complex numbers A complex number z is defined as an ordered pair z = (x, y), where x and y are a pair of real numbers. In usual notation, we write z = x + iy, where i is a symbol. The operations of

More information

2 GRAPH AND NETWORK OPTIMIZATION. E. Amaldi Introduction to Operations Research Politecnico Milano 1

2 GRAPH AND NETWORK OPTIMIZATION. E. Amaldi Introduction to Operations Research Politecnico Milano 1 2 GRAPH AND NETWORK OPTIMIZATION E. Amaldi Introduction to Operations Research Politecnico Milano 1 A variety of decision-making problems can be formulated in terms of graphs and networks Examples: - transportation

More information

Algorithms: Lecture 12. Chalmers University of Technology

Algorithms: Lecture 12. Chalmers University of Technology Algorithms: Lecture 1 Chalmers University of Technology Today s Topics Shortest Paths Network Flow Algorithms Shortest Path in a Graph Shortest Path Problem Shortest path network. Directed graph G = (V,

More information

WELL-SEPARATED PAIR DECOMPOSITION FOR THE UNIT-DISK GRAPH METRIC AND ITS APPLICATIONS

WELL-SEPARATED PAIR DECOMPOSITION FOR THE UNIT-DISK GRAPH METRIC AND ITS APPLICATIONS WELL-SEPARATED PAIR DECOMPOSITION FOR THE UNIT-DISK GRAPH METRIC AND ITS APPLICATIONS JIE GAO AND LI ZHANG Abstract. We extend the classic notion of well-separated pair decomposition [10] to the unit-disk

More information

On shredders and vertex connectivity augmentation

On shredders and vertex connectivity augmentation On shredders and vertex connectivity augmentation Gilad Liberman The Open University of Israel giladliberman@gmail.com Zeev Nutov The Open University of Israel nutov@openu.ac.il Abstract We consider the

More information

Generating p-extremal graphs

Generating p-extremal graphs Generating p-extremal graphs Derrick Stolee Department of Mathematics Department of Computer Science University of Nebraska Lincoln s-dstolee1@math.unl.edu August 2, 2011 Abstract Let f(n, p be the maximum

More information

Chapter 9: Relations Relations

Chapter 9: Relations Relations Chapter 9: Relations 9.1 - Relations Definition 1 (Relation). Let A and B be sets. A binary relation from A to B is a subset R A B, i.e., R is a set of ordered pairs where the first element from each pair

More information

On the number of cycles in a graph with restricted cycle lengths

On the number of cycles in a graph with restricted cycle lengths On the number of cycles in a graph with restricted cycle lengths Dániel Gerbner, Balázs Keszegh, Cory Palmer, Balázs Patkós arxiv:1610.03476v1 [math.co] 11 Oct 2016 October 12, 2016 Abstract Let L be a

More information

GRAPH SEARCHING, AND A MIN-MAX THEOREM FOR TREE-WIDTH. P. D. Seymour Bellcore 445 South St. Morristown, New Jersey 07960, USA. and

GRAPH SEARCHING, AND A MIN-MAX THEOREM FOR TREE-WIDTH. P. D. Seymour Bellcore 445 South St. Morristown, New Jersey 07960, USA. and GRAPH SEARCHING, AND A MIN-MAX THEOREM FOR TREE-WIDTH P. D. Seymour Bellcore 445 South St. Morristown, New Jersey 07960, USA and Robin Thomas* School of Mathematics Georgia Institute of Technology Atlanta,

More information

Chapter 1. Comparison-Sorting and Selecting in. Totally Monotone Matrices. totally monotone matrices can be found in [4], [5], [9],

Chapter 1. Comparison-Sorting and Selecting in. Totally Monotone Matrices. totally monotone matrices can be found in [4], [5], [9], Chapter 1 Comparison-Sorting and Selecting in Totally Monotone Matrices Noga Alon Yossi Azar y Abstract An mn matrix A is called totally monotone if for all i 1 < i 2 and j 1 < j 2, A[i 1; j 1] > A[i 1;

More information

Algorithms for Picture Analysis. Lecture 07: Metrics. Axioms of a Metric

Algorithms for Picture Analysis. Lecture 07: Metrics. Axioms of a Metric Axioms of a Metric Picture analysis always assumes that pictures are defined in coordinates, and we apply the Euclidean metric as the golden standard for distance (or derived, such as area) measurements.

More information

Aditya Bhaskara CS 5968/6968, Lecture 1: Introduction and Review 12 January 2016

Aditya Bhaskara CS 5968/6968, Lecture 1: Introduction and Review 12 January 2016 Lecture 1: Introduction and Review We begin with a short introduction to the course, and logistics. We then survey some basics about approximation algorithms and probability. We also introduce some of

More information

The Fundamental Group and Covering Spaces

The Fundamental Group and Covering Spaces Chapter 8 The Fundamental Group and Covering Spaces In the first seven chapters we have dealt with point-set topology. This chapter provides an introduction to algebraic topology. Algebraic topology may

More information

CMSC 451: Lecture 7 Greedy Algorithms for Scheduling Tuesday, Sep 19, 2017

CMSC 451: Lecture 7 Greedy Algorithms for Scheduling Tuesday, Sep 19, 2017 CMSC CMSC : Lecture Greedy Algorithms for Scheduling Tuesday, Sep 9, 0 Reading: Sects.. and. of KT. (Not covered in DPV.) Interval Scheduling: We continue our discussion of greedy algorithms with a number

More information

Augmenting Outerplanar Graphs to Meet Diameter Requirements

Augmenting Outerplanar Graphs to Meet Diameter Requirements Proceedings of the Eighteenth Computing: The Australasian Theory Symposium (CATS 2012), Melbourne, Australia Augmenting Outerplanar Graphs to Meet Diameter Requirements Toshimasa Ishii Department of Information

More information

Improved Approximation Bounds for Planar Point Pattern Matching

Improved Approximation Bounds for Planar Point Pattern Matching Improved Approximation Bounds for Planar Point Pattern Matching Minkyoung Cho Department of Computer Science University of Maryland College Park, Maryland, USA David M Mount Department of Computer Science

More information

CHAPTER 7. Connectedness

CHAPTER 7. Connectedness CHAPTER 7 Connectedness 7.1. Connected topological spaces Definition 7.1. A topological space (X, T X ) is said to be connected if there is no continuous surjection f : X {0, 1} where the two point set

More information

ALGEBRA. 1. Some elementary number theory 1.1. Primes and divisibility. We denote the collection of integers

ALGEBRA. 1. Some elementary number theory 1.1. Primes and divisibility. We denote the collection of integers ALGEBRA CHRISTIAN REMLING 1. Some elementary number theory 1.1. Primes and divisibility. We denote the collection of integers by Z = {..., 2, 1, 0, 1,...}. Given a, b Z, we write a b if b = ac for some

More information

Paths and cycles in extended and decomposable digraphs

Paths and cycles in extended and decomposable digraphs Paths and cycles in extended and decomposable digraphs Jørgen Bang-Jensen Gregory Gutin Department of Mathematics and Computer Science Odense University, Denmark Abstract We consider digraphs called extended

More information

On improving matchings in trees, via bounded-length augmentations 1

On improving matchings in trees, via bounded-length augmentations 1 On improving matchings in trees, via bounded-length augmentations 1 Julien Bensmail a, Valentin Garnero a, Nicolas Nisse a a Université Côte d Azur, CNRS, Inria, I3S, France Abstract Due to a classical

More information

PRIME LABELING OF SMALL TREES WITH GAUSSIAN INTEGERS. 1. Introduction

PRIME LABELING OF SMALL TREES WITH GAUSSIAN INTEGERS. 1. Introduction PRIME LABELING OF SMALL TREES WITH GAUSSIAN INTEGERS HUNTER LEHMANN AND ANDREW PARK Abstract. A graph on n vertices is said to admit a prime labeling if we can label its vertices with the first n natural

More information

Minimizing Interference in Wireless Networks. Yves Brise, ETH Zürich, Joint work with Marek, Michael, and Tobias.

Minimizing Interference in Wireless Networks. Yves Brise, ETH Zürich, Joint work with Marek, Michael, and Tobias. Minimizing Interference in Wireless Networks Yves Brise, ETH Zürich, 20100302 Joint work with Marek, Michael, and Tobias. The Problem The Problem Formal Description Input: Output: Set V of n vertices in

More information

CS6901: review of Theory of Computation and Algorithms

CS6901: review of Theory of Computation and Algorithms CS6901: review of Theory of Computation and Algorithms Any mechanically (automatically) discretely computation of problem solving contains at least three components: - problem description - computational

More information

Preliminaries and Complexity Theory

Preliminaries and Complexity Theory Preliminaries and Complexity Theory Oleksandr Romanko CAS 746 - Advanced Topics in Combinatorial Optimization McMaster University, January 16, 2006 Introduction Book structure: 2 Part I Linear Algebra

More information

Design and Analysis of Algorithms

Design and Analysis of Algorithms CSE 101, Winter 2018 Design and Analysis of Algorithms Lecture 5: Divide and Conquer (Part 2) Class URL: http://vlsicad.ucsd.edu/courses/cse101-w18/ A Lower Bound on Convex Hull Lecture 4 Task: sort the

More information

Rose-Hulman Undergraduate Mathematics Journal

Rose-Hulman Undergraduate Mathematics Journal Rose-Hulman Undergraduate Mathematics Journal Volume 17 Issue 1 Article 5 Reversing A Doodle Bryan A. Curtis Metropolitan State University of Denver Follow this and additional works at: http://scholar.rose-hulman.edu/rhumj

More information

arxiv: v4 [cs.cg] 31 Mar 2018

arxiv: v4 [cs.cg] 31 Mar 2018 New bounds for range closest-pair problems Jie Xue Yuan Li Saladi Rahul Ravi Janardan arxiv:1712.9749v4 [cs.cg] 31 Mar 218 Abstract Given a dataset S of points in R 2, the range closest-pair (RCP) problem

More information

Chapter 11. Approximation Algorithms. Slides by Kevin Wayne Pearson-Addison Wesley. All rights reserved.

Chapter 11. Approximation Algorithms. Slides by Kevin Wayne Pearson-Addison Wesley. All rights reserved. Chapter 11 Approximation Algorithms Slides by Kevin Wayne. Copyright @ 2005 Pearson-Addison Wesley. All rights reserved. 1 Approximation Algorithms Q. Suppose I need to solve an NP-hard problem. What should

More information

Disjoint paths in unions of tournaments

Disjoint paths in unions of tournaments Disjoint paths in unions of tournaments Maria Chudnovsky 1 Princeton University, Princeton, NJ 08544, USA Alex Scott Mathematical Institute, University of Oxford, Oxford OX2 6GG, UK Paul Seymour 2 Princeton

More information

Observation 4.1 G has a proper separation of order 0 if and only if G is disconnected.

Observation 4.1 G has a proper separation of order 0 if and only if G is disconnected. 4 Connectivity 2-connectivity Separation: A separation of G of order k is a pair of subgraphs (H, K) with H K = G and E(H K) = and V (H) V (K) = k. Such a separation is proper if V (H) \ V (K) and V (K)

More information

16 Chapter 3. Separation Properties, Principal Pivot Transforms, Classes... for all j 2 J is said to be a subcomplementary vector of variables for (3.

16 Chapter 3. Separation Properties, Principal Pivot Transforms, Classes... for all j 2 J is said to be a subcomplementary vector of variables for (3. Chapter 3 SEPARATION PROPERTIES, PRINCIPAL PIVOT TRANSFORMS, CLASSES OF MATRICES In this chapter we present the basic mathematical results on the LCP. Many of these results are used in later chapters to

More information

Convergence in shape of Steiner symmetrized line segments. Arthur Korneychuk

Convergence in shape of Steiner symmetrized line segments. Arthur Korneychuk Convergence in shape of Steiner symmetrized line segments by Arthur Korneychuk A thesis submitted in conformity with the requirements for the degree of Master of Science Graduate Department of Mathematics

More information

PATH PROBLEMS IN SKEW-SYMMETRIC GRAPHS ANDREW V. GOLDBERG COMPUTER SCIENCE DEPARTMENT STANFORD UNIVERSITY STANFORD, CA

PATH PROBLEMS IN SKEW-SYMMETRIC GRAPHS ANDREW V. GOLDBERG COMPUTER SCIENCE DEPARTMENT STANFORD UNIVERSITY STANFORD, CA PATH PROBLEMS IN SKEW-SYMMETRIC GRAPHS ANDREW V. GOLDBERG COMPUTER SCIENCE DEPARTMENT STANFORD UNIVERSITY STANFORD, CA 94305 GOLDBERG@CS.STANFORD.EDU AND ALEXANDER V. KARZANOV INSTITUTE FOR SYSTEMS ANALYSIS

More information

Math 42, Discrete Mathematics

Math 42, Discrete Mathematics c Fall 2018 last updated 12/05/2018 at 15:47:21 For use by students in this class only; all rights reserved. Note: some prose & some tables are taken directly from Kenneth R. Rosen, and Its Applications,

More information

Uniqueness of Optimal Cube Wrapping

Uniqueness of Optimal Cube Wrapping CCCG 014, Halifax, Nova Scotia, August 11 13, 014 Uniqueness of Optimal Cube Wrapping Qinxuan Pan EECS, MIT qinxuan@mit.edu Abstract Consider wrapping the unit cube with a square without stretching or

More information

Acyclic Digraphs arising from Complete Intersections

Acyclic Digraphs arising from Complete Intersections Acyclic Digraphs arising from Complete Intersections Walter D. Morris, Jr. George Mason University wmorris@gmu.edu July 8, 2016 Abstract We call a directed acyclic graph a CI-digraph if a certain affine

More information

Convergence of Random Walks and Conductance - Draft

Convergence of Random Walks and Conductance - Draft Graphs and Networks Lecture 0 Convergence of Random Walks and Conductance - Draft Daniel A. Spielman October, 03 0. Overview I present a bound on the rate of convergence of random walks in graphs that

More information

arxiv: v1 [cs.cg] 16 May 2011

arxiv: v1 [cs.cg] 16 May 2011 Collinearities in Kinetic Point Sets arxiv:1105.3078v1 [cs.cg] 16 May 2011 Ben D. Lund George B. Purdy Justin W. Smith Csaba D. Tóth August 24, 2018 Abstract Let P be a set of n points in the plane, each

More information

RECOVERY OF NON-LINEAR CONDUCTIVITIES FOR CIRCULAR PLANAR GRAPHS

RECOVERY OF NON-LINEAR CONDUCTIVITIES FOR CIRCULAR PLANAR GRAPHS RECOVERY OF NON-LINEAR CONDUCTIVITIES FOR CIRCULAR PLANAR GRAPHS WILL JOHNSON Abstract. We consider the problem of recovering nonlinear conductances in a circular planar graph. If the graph is critical

More information

arxiv: v1 [cs.ds] 20 Nov 2016

arxiv: v1 [cs.ds] 20 Nov 2016 Algorithmic and Hardness Results for the Hub Labeling Problem Haris Angelidakis 1, Yury Makarychev 1 and Vsevolod Oparin 2 1 Toyota Technological Institute at Chicago 2 Saint Petersburg Academic University

More information

DISCRETIZED CONFIGURATIONS AND PARTIAL PARTITIONS

DISCRETIZED CONFIGURATIONS AND PARTIAL PARTITIONS DISCRETIZED CONFIGURATIONS AND PARTIAL PARTITIONS AARON ABRAMS, DAVID GAY, AND VALERIE HOWER Abstract. We show that the discretized configuration space of k points in the n-simplex is homotopy equivalent

More information

Def. A topological space X is disconnected if it admits a non-trivial splitting: (We ll abbreviate disjoint union of two subsets A and B meaning A B =

Def. A topological space X is disconnected if it admits a non-trivial splitting: (We ll abbreviate disjoint union of two subsets A and B meaning A B = CONNECTEDNESS-Notes Def. A topological space X is disconnected if it admits a non-trivial splitting: X = A B, A B =, A, B open in X, and non-empty. (We ll abbreviate disjoint union of two subsets A and

More information

arxiv: v1 [cs.cg] 25 Dec 2018

arxiv: v1 [cs.cg] 25 Dec 2018 Evacuating Equilateral Triangles and Squares in the Face-to-Face Model Huda Chuangpishit 1, Saeed Mehrabi 2, Lata Narayanan 3, and Jaroslav Opatrny 3 1 Department of Mathematics, Ryerson University, Toronto,

More information

CS3719 Theory of Computation and Algorithms

CS3719 Theory of Computation and Algorithms CS3719 Theory of Computation and Algorithms Any mechanically (automatically) discretely computation of problem solving contains at least three components: - problem description - computational tool - analysis

More information

THE INVERSE PROBLEM FOR DIRECTED CURRENT ELECTRICAL NETWORKS

THE INVERSE PROBLEM FOR DIRECTED CURRENT ELECTRICAL NETWORKS THE INVERSE PROBLEM FOR DIRECTED CURRENT ELECTRICAL NETWORKS JOEL NISHIMURA Abstract. This paper investigates the inverse problem for the directed current networks defined by Orion Bawdon [1]. Using a

More information

Lecture 2: A Las Vegas Algorithm for finding the closest pair of points in the plane

Lecture 2: A Las Vegas Algorithm for finding the closest pair of points in the plane Randomized Algorithms Lecture 2: A Las Vegas Algorithm for finding the closest pair of points in the plane Sotiris Nikoletseas Professor CEID - ETY Course 2017-2018 Sotiris Nikoletseas, Professor Randomized

More information

Linear-Time Algorithms for Finding Tucker Submatrices and Lekkerkerker-Boland Subgraphs

Linear-Time Algorithms for Finding Tucker Submatrices and Lekkerkerker-Boland Subgraphs Linear-Time Algorithms for Finding Tucker Submatrices and Lekkerkerker-Boland Subgraphs Nathan Lindzey, Ross M. McConnell Colorado State University, Fort Collins CO 80521, USA Abstract. Tucker characterized

More information

Packing and Covering Dense Graphs

Packing and Covering Dense Graphs Packing and Covering Dense Graphs Noga Alon Yair Caro Raphael Yuster Abstract Let d be a positive integer. A graph G is called d-divisible if d divides the degree of each vertex of G. G is called nowhere

More information

Entropy-Preserving Cuttings and Space-Efficient Planar Point Location

Entropy-Preserving Cuttings and Space-Efficient Planar Point Location Appeared in ODA 2001, pp. 256-261. Entropy-Preserving Cuttings and pace-efficient Planar Point Location unil Arya Theocharis Malamatos David M. Mount Abstract Point location is the problem of preprocessing

More information

INVERSE SPANNING TREE PROBLEMS: FORMULATIONS AND ALGORITHMS

INVERSE SPANNING TREE PROBLEMS: FORMULATIONS AND ALGORITHMS INVERSE SPANNING TREE PROBLEMS: FORMULATIONS AND ALGORITHMS P. T. Sokkalingam Department of Mathematics Indian Institute of Technology, Kanpur-208 016, INDIA Ravindra K. Ahuja Dept. of Industrial & Management

More information

Arithmetic properties of the adjacency matrix of quadriculated disks

Arithmetic properties of the adjacency matrix of quadriculated disks Arithmetic properties of the adjacency matrix of quadriculated disks arxiv:math/00762v2 [mathco] 3 Aug 2003 Nicolau C Saldanha and Carlos Tomei December 22, 203 Abstract Let be a bicolored quadriculated

More information

Interference-Constrained Wireless Coverage in a Protocol Model

Interference-Constrained Wireless Coverage in a Protocol Model Interference-Constrained Wireless Coverage in a Protocol Model Prateek R. Kapadia Indian Institute of Technology, Bombay, India prateek@it.iitb.ac.in Om P. Damani Indian Institute of Technology, Bombay,

More information

Linear Algebra I. Ronald van Luijk, 2015

Linear Algebra I. Ronald van Luijk, 2015 Linear Algebra I Ronald van Luijk, 2015 With many parts from Linear Algebra I by Michael Stoll, 2007 Contents Dependencies among sections 3 Chapter 1. Euclidean space: lines and hyperplanes 5 1.1. Definition

More information

The Classification of Nonsimple Algebraic Tangles

The Classification of Nonsimple Algebraic Tangles The Classification of Nonsimple Algebraic Tangles Ying-Qing Wu 1 A tangle is a pair (B, T ), where B is a 3-ball, T is a pair of properly embedded arcs. When there is no ambiguity we will simply say that

More information

Geometric Spanners With Small Chromatic Number

Geometric Spanners With Small Chromatic Number Geometric Spanners With Small Chromatic Number Prosenjit Bose 1, Paz Carmi 1, Mathieu Couture 1, Anil Maheshwari 1, Michiel Smid 1, and Norbert Zeh 2 1 School of Computer Science, Carleton University 2

More information

Math 341: Convex Geometry. Xi Chen

Math 341: Convex Geometry. Xi Chen Math 341: Convex Geometry Xi Chen 479 Central Academic Building, University of Alberta, Edmonton, Alberta T6G 2G1, CANADA E-mail address: xichen@math.ualberta.ca CHAPTER 1 Basics 1. Euclidean Geometry

More information

Wolf-Tilo Balke Silviu Homoceanu Institut für Informationssysteme Technische Universität Braunschweig

Wolf-Tilo Balke Silviu Homoceanu Institut für Informationssysteme Technische Universität Braunschweig Multimedia Databases Wolf-Tilo Balke Silviu Homoceanu Institut für Informationssysteme Technische Universität Braunschweig http://www.ifis.cs.tu-bs.de 13 Indexes for Multimedia Data 13 Indexes for Multimedia

More information

Topological properties

Topological properties CHAPTER 4 Topological properties 1. Connectedness Definitions and examples Basic properties Connected components Connected versus path connected, again 2. Compactness Definition and first examples Topological

More information

ACO Comprehensive Exam 19 March Graph Theory

ACO Comprehensive Exam 19 March Graph Theory 1. Graph Theory Let G be a connected simple graph that is not a cycle and is not complete. Prove that there exist distinct non-adjacent vertices u, v V (G) such that the graph obtained from G by deleting

More information

Facets for Node-Capacitated Multicut Polytopes from Path-Block Cycles with Two Common Nodes

Facets for Node-Capacitated Multicut Polytopes from Path-Block Cycles with Two Common Nodes Facets for Node-Capacitated Multicut Polytopes from Path-Block Cycles with Two Common Nodes Michael M. Sørensen July 2016 Abstract Path-block-cycle inequalities are valid, and sometimes facet-defining,

More information

Definitions. Notations. Injective, Surjective and Bijective. Divides. Cartesian Product. Relations. Equivalence Relations

Definitions. Notations. Injective, Surjective and Bijective. Divides. Cartesian Product. Relations. Equivalence Relations Page 1 Definitions Tuesday, May 8, 2018 12:23 AM Notations " " means "equals, by definition" the set of all real numbers the set of integers Denote a function from a set to a set by Denote the image of

More information

Topological properties of Z p and Q p and Euclidean models

Topological properties of Z p and Q p and Euclidean models Topological properties of Z p and Q p and Euclidean models Samuel Trautwein, Esther Röder, Giorgio Barozzi November 3, 20 Topology of Q p vs Topology of R Both R and Q p are normed fields and complete

More information

Packing Arborescences

Packing Arborescences Egerváry Research Group on Combinatorial Optimization Technical reports TR-2009-04. Published by the Egerváry Research Group, Pázmány P. sétány 1/C, H1117, Budapest, Hungary. Web site: www.cs.elte.hu/egres.

More information

CS 6820 Fall 2014 Lectures, October 3-20, 2014

CS 6820 Fall 2014 Lectures, October 3-20, 2014 Analysis of Algorithms Linear Programming Notes CS 6820 Fall 2014 Lectures, October 3-20, 2014 1 Linear programming The linear programming (LP) problem is the following optimization problem. We are given

More information

arxiv: v1 [math.co] 23 Feb 2015

arxiv: v1 [math.co] 23 Feb 2015 Catching a mouse on a tree Vytautas Gruslys Arès Méroueh arxiv:1502.06591v1 [math.co] 23 Feb 2015 February 24, 2015 Abstract In this paper we consider a pursuit-evasion game on a graph. A team of cats,

More information

THE MAXIMAL SUBGROUPS AND THE COMPLEXITY OF THE FLOW SEMIGROUP OF FINITE (DI)GRAPHS

THE MAXIMAL SUBGROUPS AND THE COMPLEXITY OF THE FLOW SEMIGROUP OF FINITE (DI)GRAPHS THE MAXIMAL SUBGROUPS AND THE COMPLEXITY OF THE FLOW SEMIGROUP OF FINITE (DI)GRAPHS GÁBOR HORVÁTH, CHRYSTOPHER L. NEHANIV, AND KÁROLY PODOSKI Dedicated to John Rhodes on the occasion of his 80th birthday.

More information

Claw-free Graphs. III. Sparse decomposition

Claw-free Graphs. III. Sparse decomposition Claw-free Graphs. III. Sparse decomposition Maria Chudnovsky 1 and Paul Seymour Princeton University, Princeton NJ 08544 October 14, 003; revised May 8, 004 1 This research was conducted while the author

More information

Robust Network Codes for Unicast Connections: A Case Study

Robust Network Codes for Unicast Connections: A Case Study Robust Network Codes for Unicast Connections: A Case Study Salim Y. El Rouayheb, Alex Sprintson, and Costas Georghiades Department of Electrical and Computer Engineering Texas A&M University College Station,

More information

Classical Complexity and Fixed-Parameter Tractability of Simultaneous Consecutive Ones Submatrix & Editing Problems

Classical Complexity and Fixed-Parameter Tractability of Simultaneous Consecutive Ones Submatrix & Editing Problems Classical Complexity and Fixed-Parameter Tractability of Simultaneous Consecutive Ones Submatrix & Editing Problems Rani M. R, Mohith Jagalmohanan, R. Subashini Binary matrices having simultaneous consecutive

More information

IS 709/809: Computational Methods in IS Research Fall Exam Review

IS 709/809: Computational Methods in IS Research Fall Exam Review IS 709/809: Computational Methods in IS Research Fall 2017 Exam Review Nirmalya Roy Department of Information Systems University of Maryland Baltimore County www.umbc.edu Exam When: Tuesday (11/28) 7:10pm

More information

Learning Large-Alphabet and Analog Circuits with Value Injection Queries

Learning Large-Alphabet and Analog Circuits with Value Injection Queries Learning Large-Alphabet and Analog Circuits with Value Injection Queries Dana Angluin 1 James Aspnes 1, Jiang Chen 2, Lev Reyzin 1,, 1 Computer Science Department, Yale University {angluin,aspnes}@cs.yale.edu,

More information

Linear Algebra (part 1) : Vector Spaces (by Evan Dummit, 2017, v. 1.07) 1.1 The Formal Denition of a Vector Space

Linear Algebra (part 1) : Vector Spaces (by Evan Dummit, 2017, v. 1.07) 1.1 The Formal Denition of a Vector Space Linear Algebra (part 1) : Vector Spaces (by Evan Dummit, 2017, v. 1.07) Contents 1 Vector Spaces 1 1.1 The Formal Denition of a Vector Space.................................. 1 1.2 Subspaces...................................................

More information

Induced Subgraph Isomorphism on proper interval and bipartite permutation graphs

Induced Subgraph Isomorphism on proper interval and bipartite permutation graphs Induced Subgraph Isomorphism on proper interval and bipartite permutation graphs Pinar Heggernes Pim van t Hof Daniel Meister Yngve Villanger Abstract Given two graphs G and H as input, the Induced Subgraph

More information

A Decidable Class of Planar Linear Hybrid Systems

A Decidable Class of Planar Linear Hybrid Systems A Decidable Class of Planar Linear Hybrid Systems Pavithra Prabhakar, Vladimeros Vladimerou, Mahesh Viswanathan, and Geir E. Dullerud University of Illinois at Urbana-Champaign. Abstract. The paper shows

More information

Math 541 Fall 2008 Connectivity Transition from Math 453/503 to Math 541 Ross E. Staffeldt-August 2008

Math 541 Fall 2008 Connectivity Transition from Math 453/503 to Math 541 Ross E. Staffeldt-August 2008 Math 541 Fall 2008 Connectivity Transition from Math 453/503 to Math 541 Ross E. Staffeldt-August 2008 Closed sets We have been operating at a fundamental level at which a topological space is a set together

More information

Voronoi Diagrams for Oriented Spheres

Voronoi Diagrams for Oriented Spheres Voronoi Diagrams for Oriented Spheres F. Aurenhammer J. Wallner University of Technology Graz, Austria auren@igi.tugraz.at j.wallner@tugraz.at M. Peternell H. Pottmann University of Technology Vienna,

More information

MINIMALLY NON-PFAFFIAN GRAPHS

MINIMALLY NON-PFAFFIAN GRAPHS MINIMALLY NON-PFAFFIAN GRAPHS SERGUEI NORINE AND ROBIN THOMAS Abstract. We consider the question of characterizing Pfaffian graphs. We exhibit an infinite family of non-pfaffian graphs minimal with respect

More information

Even Cycles in Hypergraphs.

Even Cycles in Hypergraphs. Even Cycles in Hypergraphs. Alexandr Kostochka Jacques Verstraëte Abstract A cycle in a hypergraph A is an alternating cyclic sequence A 0, v 0, A 1, v 1,..., A k 1, v k 1, A 0 of distinct edges A i and

More information

Discrete Mathematics

Discrete Mathematics Discrete Mathematics 309 (009) 3811 380 Contents lists available at ScienceDirect Discrete Mathematics journal homepage: www.elsevier.com/locate/disc Disjoint hamiltonian cycles in bipartite graphs Michael

More information

Chapter 11. Approximation Algorithms. Slides by Kevin Wayne Pearson-Addison Wesley. All rights reserved.

Chapter 11. Approximation Algorithms. Slides by Kevin Wayne Pearson-Addison Wesley. All rights reserved. Chapter 11 Approximation Algorithms Slides by Kevin Wayne. Copyright @ 2005 Pearson-Addison Wesley. All rights reserved. 1 P and NP P: The family of problems that can be solved quickly in polynomial time.

More information

Graph Theory. Thomas Bloom. February 6, 2015

Graph Theory. Thomas Bloom. February 6, 2015 Graph Theory Thomas Bloom February 6, 2015 1 Lecture 1 Introduction A graph (for the purposes of these lectures) is a finite set of vertices, some of which are connected by a single edge. Most importantly,

More information

Analysis of Algorithms. Outline. Single Source Shortest Path. Andres Mendez-Vazquez. November 9, Notes. Notes

Analysis of Algorithms. Outline. Single Source Shortest Path. Andres Mendez-Vazquez. November 9, Notes. Notes Analysis of Algorithms Single Source Shortest Path Andres Mendez-Vazquez November 9, 01 1 / 108 Outline 1 Introduction Introduction and Similar Problems General Results Optimal Substructure Properties

More information

Eilenberg-Steenrod properties. (Hatcher, 2.1, 2.3, 3.1; Conlon, 2.6, 8.1, )

Eilenberg-Steenrod properties. (Hatcher, 2.1, 2.3, 3.1; Conlon, 2.6, 8.1, ) II.3 : Eilenberg-Steenrod properties (Hatcher, 2.1, 2.3, 3.1; Conlon, 2.6, 8.1, 8.3 8.5 Definition. Let U be an open subset of R n for some n. The de Rham cohomology groups (U are the cohomology groups

More information

MATH 215B. SOLUTIONS TO HOMEWORK (6 marks) Construct a path connected space X such that π 1 (X, x 0 ) = D 4, the dihedral group with 8 elements.

MATH 215B. SOLUTIONS TO HOMEWORK (6 marks) Construct a path connected space X such that π 1 (X, x 0 ) = D 4, the dihedral group with 8 elements. MATH 215B. SOLUTIONS TO HOMEWORK 2 1. (6 marks) Construct a path connected space X such that π 1 (X, x 0 ) = D 4, the dihedral group with 8 elements. Solution A presentation of D 4 is a, b a 4 = b 2 =

More information