J.W. Green and K. J. Supowit, `Simulated Annealing without Rejected Moves', in Digest. of Intl. Conference on Computer Design, pp. 658-663, Oct. 1984.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....stages of a search, when some low predetermined acceptance ratio is reached, then to proceed with another algorithm. A rejectionless method was developed, similar in spirit to this method, but more methodical, and yielding a search time not dependent on the acceptance ratio or temperature [81]. Acceptance criteria are biased according to information being gathered on the cost function during the search, maintaining detailed balance throughout the search. They suggest using standard SA until some low acceptance criteria is reached, then to finish annealing using their method. Some ....

J.A. Greene and K.J. Supowit, Simulated annealing without rejected moves, IEEE Trans. Comput. -Aided Design CAD-5 (1), 221-228 (1986).

....new sequence up to that point and discard (flush) the model. A new Markov model is started again and the process is continued until the original sequence is completely processed. For the generation phase itself, we use a modified version of the dynamic weighted selection algorithm described in [22]. In general, by alternating the generation and flush phases in the DMC procedure, the complexity of the model can be effectively handled. The only remaining issue is to determine how many vectors must be generated inside each macrostate before a transition to another macrostate is performed. If a ....

....sequence and simultaneously build the basic tree DMT ; during this process, the frequency counts on DMT s edges are dynamically updated. During this process, the order of the source is also determined. The generation step is done using a modified version of the weighted selection algorithm [22]. Finally, a validation step is included in the strategy; we have used an in house gate level logic simulator developed under SIS. The total power consumption of some sequential benchmarks has been measured for the initial and the compacted sequences, making it possible to assess the effectiveness ....

J. W. Green and K. J. Supowit, "Simulated annealing without rejected moves," in Dig. Int. Conf. Computer Design, Oct. 1984, pp. 658--663.

....at some temperature close to zero. Cooling schedules are discussed in detail in Chapter 2. In addition to alternative cooling schedules, several other approaches to accelerating SA on a single processor have been investigated. These include alternative neighbor generation acceptance strategies [37], noisy cost evaluation [39] and optimal finite time temperature scheduling [9, 42, 43, 94] In addition to these uni processor approaches to acceleration, the parallelization of the paradigm has received significant attention in the literature [2, 4, 12, 13, 22, 27, 49, 50, 53, 58, 87, 89, 103] ....

....but again no formal statements are made as to the effect on convergence. This remains an open problem. 33 2.5. 2 Alternative Transition Generation The second method of SA acceleration to be examined here is the alternative transition generation acceptance mechanism of Greene and Supowit [37]. First, Green and Supowit define a supplementary function based on the traditional acceptance function of Equation (2.7) 2.48) Next, they propose an alternative generation mechanism to the standard generation matrix G described by Equations (2.27) and (2.32) Greene and Supowit define their ....

[Article contains additional citation context not shown here]

J.W. Greene and K.J. Supowit, "Simulated Annealing Without Rejected Moves," IEEE Trans. CADICS, vol. 5, 221-228, 1986.

.... There is a closer resemblance between EO and algorithms such as GSAT (for satisfiability) that choose, at each update step, the move resulting in the best subsequent outcome whether or not that outcome is an improvement over the current solution [32] Also, versions of SA have been proposed [17,31] that enforce equilibrium dynamics by ranking local moves according to anticipated outcome, and then choosing them Fig. 2. Evolution of the cutsize during an extremal optimization run on the N = 500 geometric graph with C = 5 (see Fig. 1) The shaded area marks the range of cutsizes explored ....

J.W. Greene, K.J. Supowit, Simulated annealing without rejected moves, IEEE Trans. Computer-Aided Design CAD-5 (1986) 221--228.

....generate the new sequence up to that point and discard (flush) the model. A new Markov model is started again and the process is continued until the original sequence is completely processed. For the generation phase itself, we use a modified version of the dynamic weighted selection algorithm [12]. The pseudocode for the generation phase and detailed examples involving flushes are given in [14] In general, by alternating the generation and flush phases in the DMC procedure, the complexity of the model can be effectively handled. To see how the flushing technique affects the accuracy, ....

J.W. Green and K.J. Supowit, `Simulated Annealing without Rejected Moves', in Digest. of Intl. Conference on Computer Design, pp. 658-663, Oct. 1984.

....method, and the net based partitioning method. Some of the best known iterative improvement based partitioning methods include the Kernighan Lin (KL) algorithm [22] the Fiduccia Mattheyses (FM) algorithm [15] the FM algorithm with look ahead scheme [24] and the simulated annealing approach [23, 17]. The analytical methods include both the use of a linear placement formulation with the quadratic objective function, which is solved by computing the second smallest eigenvector of the Laplacian matrix of the given network [14, 2, 4, 18] and the use of the linear placement formulation with a ....

J. Greene and K. Supowit, "Simulated Annealing without Rejected Moves," Proc. Int'l Conf. on Computer Design, pp. 658-663, 1984.

....if the temperature is low. The simulated annealing algorithm was modied so that if no move is accepted for a given time, the algorithm examines the whole neighborhood and uses this information to calculate the probabilities for each move solution to be the next accepted move in the same way as in (Greene and Supowit, 1986), except that we used a simpler (and less eOEcient) algorithm to select the move after all the cost changes in the neighborhood had been calculated. However, the gain obtained by using the eOEcient selection algorithm in (Greene and Supowit, 1986) would be negligible, because even the simple ....

....to be the next accepted move in the same way as in (Greene and Supowit, 1986) except that we used a simpler (and less eOEcient) algorithm to select the move after all the cost changes in the neighborhood had been calculated. However, the gain obtained by using the eOEcient selection algorithm in (Greene and Supowit, 1986) would be negligible, because even the simple selection algorithm uses several orders of magnitude less time than the cost change calculations. Because the range of temperatures where the search works well is rather wide, it is tempting to go to the other extreme and use constant temperature ....

Greene, J. W. and Supowit, K. J. (1986). Simulated annealing without rejected moves, IEEE Trans. CAD 5: 221228.

....produces high quality approximate solutions for these problems. SA has only one significant disadvantage its typically very long computation time. There has been considerable effort aimed at speeding up SA. Most of this work has concentrated on the development of faster cooling schedules [2, 8, 13, 19, 20, 25]. Another approach is two stage simulated annealing (TSSA) 9, 10, 14, 28, 29] For TSSA, a faster heuristic algorithm is used to replace the SA actions occurring at the highest temperatures in the cooling schedule. The heuristic is then followed by a conventional SA approach initiated at a lower ....

....value, the acceptance function, the length of the Markov chain at each temperature, the temperature decrement rule, and the stopping conditions. Collectively these implementation parameters are known as the cooling schedule [16] There are many proposed cooling schedules present in the literature [2, 4, 8, 13, 17, 19, 20, 25, 30]. For the purpose of this paper, we concern ourselves with traditional cooling schedules that conform to the Markov chain model described above. The reason Simulated Annealing( initialize(i, t) i best = i; do do j = perturb(i) Dc ij = c(j) c(i) if ( Dc ij 0) random( ....

J.W. Greene and K.J. Supowit, "Simulated Annealing Without Rejected Moves," IEEE Trans. CADICS, vol. 5, 221-228, 1986.

....birth and death as a sampler, and generally simpler to implement. On the other hand, optimization algorithms based on continuous time processes [3] seem to be much more efficient than those based on standard discrete time processes, such as simulated annealing, for certain optimization problems [4]. The discrete time processes in general use suffer from too high a rejection rate in the low temperature domain, whilst algorithms set in continuous time simulate an imbeded chain of transitions and therefore always get an update . Such algorithms have been used in the physics and optimization ....

JW Greene and KJ Supowit. Simulated annealing without rejected moves. IEEE transactions on computeraided design of integrated circuits and systems, 5(1):221--228, 1986.

....variables, T j = c i j ; j = 1; n, and a position of the transparent nodes sequence l. A neighboring configuration is given by an assignation in which the value of T r K (l) can be modified and the current place is K(l) The simulated annealing algorithm follows Green and Supowit, [23] modification to improve the algorithm efficiency. According to it, if we are in configuration C if we can calculate the objective PC 0 (X j = uje) of all neighbouring configurations of and choose configuration C 0 with a probability proportional to e (P C 0 (X j =uje) GammaP C (X j =uje) t . ....

J.W. Greene and K.J. Supowit. Simulated annealing without rejected moves. In Proc. IEEE Int. Conference on Computer Design, pages 658-- 663, Port Chester, 1984.

....and the net based partitioning method. Some of the best known iterative improvement based partitioning methods include the Kernighan Lin (KL) algorithm [KeLi70] the Fiduccia and Mattheyses (FM) algorithm [FiMa82] the FM algorithm with lookahead scheme [Kr83] and the simulated annealing approach [KiGV83, GrSu84]. The analytical methods use a linear placement formulation with either the quadratic wirelength objective function, which is solved by computing the second smallest eigenvector of the Laplacian matrix of the given circuit [Ba82, Bo87, DoHo73, HaKa92, AlYa95] or a linear wirelength objective ....

Jonathan Greene, Kenneth Supowit, "Simulated Annealing without Rejected Moves", Proc. Int'l Conf. on Computer-Aided Design, pp. 658-663, 1984

....randomly, and then moves cells between partitions to improve the solution. Some of the best known methods include the Kernighan Lin (KL) algorithm [KL70] the Fiduccia Mattheyses (FM) algorithm [FM82] Krishnamurthy s lookahead scheme [Kri84] and the simulated annealing based approach [KGV83, GS84] The analytical method uses a linear placement formulation with either (i) the quadratic wire length objective function solved by computing the second smallest eigenvector of the Laplacian matrix of the given circuit [Bar82, Bop87, DH73, HK92, AY95] or (ii) the linear wire length objective ....

J. Greene and K. Supowit. "Simulated annealing without rejected moves". In Proc. Int'l Conf. on Computer-Aided Design, pages 658--663, 1984.

....results and in Section 5 we present our conclusions and area for future work. 2. SIMULATED ANNEALING AND REGRESSION A number of definitions are introduced below for use in presenting the overall approach. Page 3 2. 1 Simulated Annealing There are many versions of the simulated annealing [14, 17 19]. We will outline the basic framework of the algorithm below. If S represents a particular configuration of the placement, T is the current temperature of the annealing, S C returns the cost of configuration S, m is a move (in our case, a two cell swap) S is the new configuration due to the ....

J. W. Greene and K. J. Supowit, "Simulated annealing without rejected moves", International Conference on Computer Design, 1984, 658-63

....is an active area of computing science research. There are many different techniques for solving such problems. A good general technique is simulated annealing [4] Rejectionless annealing extends conventional simulated annealing to improve the solutions obtained at the cost of increased run time [2]. Parallel computing can greatly assist rejectionless annealing by improving its run time efficiency. It will demonstrated that parallel rejectionless annealing algorithm is an efficient parallel algorithm for combinatorial optimisation problems. It will be shown that the algorithm scales well ....

....1: Metropolis Algorithm 1953. The algorithm has become a useful computational model for the solution of combinatorial optimisation problems. However at low temperatures the algorithm becomes less efficient, because at lower temperatures less acceptable changes are generated. Greene and Supowit [2] devised the rejectionless form of the annealing algorithm to overcome this problem at low temperatures. Rejectionless annealing unlike conventional simulated annealing, randomly selects a possible change in state from the set of all possible changes, weighted by the probability of the change. ....

J. Greene and K. Supowit. Simulated annealing without rejected moves. In IEEE International Conference on Computer Design, pages 658-- 663, New York, 1984.

....phase is driven by the user specified compaction parameter ratio that is, in order to generate a total of m = n ratio vectors, we have to keep the same compaction ratio for every dynamically grown Markov model. For generation, we use a modified version of the dynamic weighted selection algorithm [20]. In that approach, a similar structure with DMT 0 is built; more precisely, a full tree having on the leaves the symbols that need to be generated. The counts on the edges are dynamically decreased and the symbols are generated according to their probability distribution. We use this strategy ....

J.W.Green and K.J.Supowit, `Simulated Annealing without Rejected Moves', in Digest. of Intl. Conference on Computer Design, pp. 658663, Oct. 1984.

.... (a) b) 14 Fig.8 Once we built the Markov tree in Fig.8, we start the procedure generate seq with parameter ratio = 2 and generate a subset of 8 vectors which best approximate the original sequence. To this effect, we use a modified version of the dynamic weighted selection algorithm [20]. In that approach, a similar structure with DMT 0 is built; more precisely, a full tree having on the leaves the symbols that need to be generated. The counts on the edges are dynamically updated and the symbols are generated according to their probability distribution. For this, a single random ....

....the frequency counts on DMT 1 s edges are dynamically updated. The next step in Fig.10 does the actual generation of the output sequence (of length m) As explained in Section 4, to generate the new sequence we use a modified version of the dynamic weighted selection algorithm presented in [20]. If the initial sequence has the length n and the new generated sequence has the length m n then the outcome of this process is a compacted sequence, equivalent to the initial one as far as total power consumption is concerned; we say that a compaction ratio of r = n m was achieved. Finally, a ....

J.W.Green and K.J.Supowit, `Simulated Annealing without Rejected Moves', in Digest. of Intl. Conference on Computer Design, pp. 658-663, Oct. 1984

....phase is driven by the userspecified compaction parameter ratio that is, in order to generate a total of m = n ratio vectors, we keep the same compaction ratio for every dynamically grown Markov model. For the generation step, we use a modified version of the dynamic weighted selection algorithm [20]. In that approach, a similar structure with DMT 0 is built; more precisely, a full tree having on the leaves the symbols that need to be generated. The counts on the edges are dynamically changed and the symbols are generated according to their probability distribution. For this, a single random ....

J.W.Green and K.J.Supowit, `Simulated Annealing without Rejected Moves', in Digest. of Intl. Conference on Computer Design, pp. 658-663, Oct. 1984.

....Gelatt and Vecchi [9] With this procedure, on each step the temperature is decreased according to the formula: t i 1 = ff:t i , where ff is a given constant. This will be the procedure used in this paper. Another modification that can improve the efficiency was proposed by Green and Supowit, [8]. According to it, if we can calculate the cost of all neighbouring configurations of N(x) then instead of randomly choosing a configuration of N(x) and accepting it according to above procedure, it is better to choose a configuration, y, from N(x) with a probability proportional to e ....

Greene J.W., K.J. Supowit (1984) Simulated annealing without rejected moves. Proc. IEEE Int. Conference on Computer Design, Port Chester, 658-663.

....and the net based partitioning method. Some of the best known iterative improvement based partitioning methods include the Kernighan Lin (KL) algorithm [KeLi70] the FiducciaMattheyses (FM) algorithm [FiMa82] the FM algorithm with look ahead scheme [Kr84] and the simulated annealing approach[KiGV83, GrSu84]. The analytical methods include both the use of a linear placement formulation with the 2quadratic objective function, which is solved by computing the second smallest eigenvector of the Laplacian matrix of the given network [DoHo73, Ba82, Bo87, HaKa91] and the use of the linear placement ....

Greene, J. and K. Supowit, "Simulated Annealing without Rejected Moves," Proc. Int'l Conf. on Computer Designs, pp. 658-663, 1984.

No context found.

J.W. Green and K. J. Supowit, `Simulated Annealing without Rejected Moves', in Digest. of Intl. Conference on Computer Design, pp. 658-663, Oct. 1984.

No context found.

J.W.Green and K.J.Supowit, `Simulated Annealing without Rejected Moves', in Digest. of Intl. Conference on Computer Design, pp. 658663, Oct. 1984

No context found.

J.W. Greene and K. J. Supowit, Simulated annealing without rejected moves, IEEE Transactions on Computer-aided Design, CAD-5, January 1986, 221--228.

No context found.

J.W. Greene and K. J. Supowit. Simulated annealing without rejected moves, IEEE Transactions on Computer-aided Design, vol CAD-5, 1, January 1986, 221-228.

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