J. Zhang, Personal communication. 11  Home/Search   Document Not in Database   Summary   Related Articles   Check

....f i is concave, kf i (j) f i (kj) so we are only overestimating the cost. The above proposition together with the soft capacitated facility location algorithm of Mahdian et al. 20] yields a 2.89 approximation algorithm for the concave facility location problem. It was noticed by J. Zhang [27] that it is in fact possible to reduce the concave facility location problem to the uncapacitated facility location problem. This reduction, together with the algorithm of Mahdian et al. 20] gives an algorithm with an approximation factor of 1.52 for the concave facility location problem with a ....

J. Zhang, Personal communication. 11

....if for each pixel in the output image, its corresponding pixels in HOM have the same or less opacity . The exact condition for HOM being conservative is to store a floating point coverage percentage per pixel. At that point, HOM becomes conservative to the same extent as object space algorithms [122]. Since authors use a limited base resolution (256x256) high quality antialiasing capabilities are required in the hardware. Depth DEB functionality has a strong resemblance with the Z buffer. Whereas the Z buffer is initialized to the farthest possible value, DEB is initialized to the nearest ....

Hansong Zhang. Personal communication, 2000.

....than a certain percentage of sentences in common. The second difficulty is one of complexity: the only algorithm known to us to compute the best matching as defined above (based on post processing the output of the algorithm in [ZS89] runs in time O(n 2 ) where n is the number of tree nodes [Zha95] To solve the first difficulty, we restrict the set of 4.3. FINDING GOOD MATCHINGS 53 matchings we consider by introducing stronger matching criteria, as described below. These criteria also permit us to design efficient algorithms for matching. In the rest of this section, we describe some ....

K. Zhang. Personal communication, May 1995.

....components is trained on one half of the training set. Further training is done using the other half of the training set to learn to combine results from the three different components, using a second neural network. The best results they report are 66.4 for a training set size of 96 proteins [Zhang91] The major difference between these learning approaches and FSKBANN is that only FSKBANN incorporates domain knowledge. FSKBANN also differs in that the neural networks used in our studies uses state information, unlike the above approaches. 6.2. Methods of Representing State Information in ....

X. Zhang, Personal communication, 1991.

....Another difference is that in RSVP, the number of sources sending to a destination is limited by active filtering within the network. In contrast, channel groups specify relationships that does not require actions to be taken by the network. In the predicted service [ClShZh92] JaShZh92] [Zhang93], classes are used for QoS specification. Also, considerable work has been done by Van Jacobson and Sally Floyd in partitioning the network resources between paying organizational entities and or application classes [Floyd93] At first glance, the class mechanism looks like a special form of group ....

L. Zhang, personal communication (Jan-Mar 1993)

....troublesome, interpolation operator behaves well in both two and three dimensional space in our applications. For technical reasons, we assume that the coarse mesh is quasi uniform, however, no such assumption is needed for the fine mesh. Other recent developments along this line can be found in [9, 10, 11, 19, 20]. Some theory and experiments with the pointwise interpolation in the context of non nested multigrid methods can be found in [1, 8, 16, 17] and references therein. In [14] Dryja and Widlund developed a general theory for Schwarz type algorithms which has a convergence rate characterized by the ....

....VHc ae H 1 0( Omega Gamma8 is indeed bounded. In a trivial case, when VHc ae V h then (i) of Lemma 2.1 holds as an equality with C = 1, and (ii) holds with C = 0. The elementary proof provided in this section applies to both d = 2 and 3. Several different proofs have been obtained recently, see [9, 11, 19, 20]. We begin with part (i) of Lemma 2.1. The essential step is to establish the estimate j Pi h uj 2 H 1 (k) C iX jruj j 2 2 j jkj; 8u 2 VHc ; 13) where k 2 Omega h , and the summation is taken over all 2 Omega Hc that have non empty intersection with k. We note that ruj is a ....

X. Zhang, Personal communication.

....be routed in at most 3n steps. Roberts, Symvonis and Zhang [20] reduced the number of steps to at most 2:3n. Furthermore, for the special cases of bounded degree trees and complete d ary trees of n nodes, they showed that routing terminates after 2n o(n) and n o(n) steps, respectively. Zhang [21] subsequently reduced the number of steps required to route a permutation on an arbitrary tree to 2n. The only work related to on line routing on trees consists of the study of sorting on linear arrays based on the odd even transposition method [11] see also [1,16] The odd even transposition ....

L. Zhang. Personal communication, 1996.

....than a certain percentage of sentences in common. The second difficulty is one of complexity: the only algorithm known to us to compute the best matching as defined above (based on post processing the output of the algorithm in [ZS89] runs in time O(n 2 ) where n is the number of tree nodes [Zha95] To solve the first difficulty, we restrict the set of matchings we consider by introducing stronger matching criteria, as described below. These criteria also permit us to design efficient algorithms for matching. In the rest of this section, we describe some matching criteria for keyless data, ....

K. Zhang. Personal communication, May 1995.

....for I h = Pi h , in three dimensions (27) and (28) are not true for all u 2 H 1 ( Omega Gamma2 Fortunately, they are true in the general finite element spaces. We state this fact in the following lemma. Several alternative proofs for this result exist, see, for instance, Cai [4] Zhang [26] and Widlund [23] Lemma 4.1. Assume that T h and T H are both shape regular, not necessarily quasiuniform, and W h and W H are any two corresponding finite element spaces consisting of continuous piecewise polynomials defined on Omega and Omega H , respectively. Furthermore, we assume ....

X. Zhang. Personal Communication.

....than a certain percentage of sentences in common. The second difficulty is one of complexity: the only algorithm known to us to compute the best matching as defined above (based on post processing the output of the algorithm in [ZS89] runs in time O(n 2 ) where n is the number of tree nodes [Zha95] To solve the first difficulty, we restrict the set of matchings we consider by introducing stronger matching criteria, as described below. These criteria also permit us to design efficient algorithms for matching. In the rest of this section, we describe some matching criteria for keyless data, ....

K. Zhang. Personal communication, May 1995.

....the finite model finder transforms a set of first order clauses into ground clauses such that the ground clauses are satisfiable if and only if the first order clause set has a model of size n. The ground clauses are tested for satisfiability using a very fast Davis Putnam procedure developed by H Zhang [Zha] The model finding algorithm starts with a domain of size one and proceeds on to the next size if the ground clause set is unsatisfiable, or stops with the model if it is satisfiable. We have developed some preprocessing techniques that makes our system efficient. For subsumption checking ....

....finder transforms a set of first order clauses into ground clauses such that the ground clauses are satisfiable if and only if the first order clause set has a model of size n. The ground clauses are tested for satisfiability using a very fast Davis Putnam procedure developed by H Zhang [Zha] The model finding algorithm starts with a domain of size one and proceeds on to the next size if the ground clause set is unsatisfiable, or stops with the model if it is satisfiable. We have developed some preprocessing techniques that makes our system efficient. For subsumption checking we ....

[Article contains additional citation context not shown here]

Hantao Zhang. Personal Communication.

....working through three examples, we have demonstrated the generality of the kinetization procedure, which transforms a static algorithm into its kinetic counterpart. The algorithms described in this paper have recently been implemented by Craig Silverstein (convex hull) and Li Zhang (closest pair) [23, 24]. In both cases, finite precision arithmetic generated errors in the exact sequencing of events, which were due to two main causes. The first was keeping a global clock for event times as more and more bits are required to record the current time, precision is lost as the simulation proceeds. A ....

....algorithm in Section 3. Experiments on random inputs showed that our kinetic convex hull algorithm has an overhead of internal events that is of the same order as the number of exterPage nal events [23] whereas our kinetic closest pair algorithm always processes Theta(n 2 ) internal events [24]. Hence, ideally, the measure of efficiency should not compare the worst case number of internal events to the worst case number of external events, but the worst case ratio of the actual number of internal events to the actual number of external events for any flight plan. It appears much more ....

L. Zhang. Personal communication. 1996.

No context found.

W. Zhang, personal communication, 18 September, 1994.

No context found.

L. Zhang. Personal communication. 1996.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC