An Efficient Sequential SAT Solver With Improved Search Strategies

Size: px
Start display at page:

Download "An Efficient Sequential SAT Solver With Improved Search Strategies"

Transcription

1 An Effiient Sequential SAT Solver With Improved Searh Strategies F. Lu, M.K. Iyer, G. Parthasarathy, L.-C. Wang, and K.-T. Cheng K.C. Chen Department of ECE, Design Tehnology University of California at Santa Barara Cadene Corporation Astrat A sequential SAT solver Satori [1] was reently proposed as an alternative to ominational SAT in verifiation appliations. This paper desries the design of Seq-SAT an effiient sequential SAT solver with improved searh strategies over Satori. The major improvements inlude (1) a new and etter heuristi for minimizing the set of assignments to state variales, (2) a new priority-ased searh strategy and a flexile sequential searh framework whih integrates different searh strategies, and (3) a deision variale seletion heuristi more suitale for solving the sequential prolems. We present experimental results to demonstrate that our sequential SAT solver an ahieve orders-of-magnitude speedup over Satori. We plan to release the soure ode of Seq-SAT along with this paper. I. Introdution Boolean SAT finds appliations in many areas of iruit design and verifiation suh as Bounded Model Cheking [2, 16], Unounded Model Cheking [13, 14], Redundany Identifiation, Equivalene Cheking [3], Preimage Calulation [15], et. State-of-the-art SAT algorithms, as implemented in tools suh as ZCHAFF [4],BERKMIN [5], and C-SAT [6] have demonstrated that very hard SAT prolems an now e solved in reasonale time. Bounded sequential searh using SAT has een shown to e very effetive in model heking. However, its major disadvantage is its lak of ompleteness in general sequential searh. A sequential SAT solver Satori was proposed in [1], whih utilizes omined ATPG and SAT tehniques to realize a sequential SAT solver y retaining the effiieny of Boolean SAT and eing omplete in the searh. Given a sequential iruit, we assume that the iruit follows the Huffman synhronous sequential iruit model as illustrated in Figure 1. Sequential SAT (or sequential justifiation) is the prolem of finding an ordered sequene of input assignments to a sequential iruit, suh that a desired ojetive is satisfied, or proving that no suh sequene exists. A desired ojetive an e a olletion of signal value onstraints suh as that a primary output is 1. In this paper, we fous on the sequential prolems that initial states are given. PO PO PO PI PI PI PPI Cominational Logi Cominational Logi PPO PPI PPO PPI Figure 1: Bakward time-frame expansion Cominational Logi In sequential SAT, a sequential iruit is oneptually unfolded into multiple opies through akward time-frame expansion (Figure 1). In eah time-frame, the iruit eomes ominational and hene, a ominational SAT solver an e applied. In eah time-frame, a PPO state element suh as a flip-flop is translated into two orresponding signals: a pseudo primary input (PPI) and a pseudo primary output (PPO). The initial state is speified at the PPIs in time-frame 0. The ojetive is speified at the signals in time-frame n (the last timeframe, where n is unknown efore solving the prolem). While solving in an intermediate time-frames i (0< i < n 1), intermediate state solutions are produed at the PPIs and they eome intermediate ojetives for further justifiation in the previous time-frames i 1. Seq-SAT has three major improvements over Satori: State redution As pointed out in the previous work [1], minimizing the set of assignments to the state variales, alled state redution, is a ritial step for improving the effiieny of sequential searh. Seq-SAT performs state redution not only when an intermediate solution is produed ut also when an intermediate state ojetive is proved unsatisfiale. In the first ase, it employs an effiient two-step state redution algorithm to otain a smaller state lause. The algorithm follows a similar proess as the D-algorithm [8] originally proposed for ATPG and utilizes 3-value simulation. For the seond ase, a different onflit analysis proedure rather than UIP ased onflit analysis[12] is applied to derive a smaller state lause. Flexile sequential searh framework Sequential searh an e arried out in two strategies: depth-first or readth-first. Inthe depth-first strategy, a SAT algorithm expands the searh to a new timeframe whenever a solution state is identified for the urrent timeframe. If the akward justifiation fails in timeframe i 1 and no state solution an e found, it then aktraks to timeframe i and selet another state solution of timeframe i to ontinue the justifiation. In the readth-first strategy, the state solutions in the urrent timeframe i are exhausted efore expanding into a new timeframe i 1. In this paper, we provide a flexile sequential searh framework in whih these two searh strategies an e integrated. We introdue a priority-ased searh strategy that an omine oth depthfirst searh and readth-first searh and demonstrate its effiieny over other searh strategies. Deision variale seletion heuristi InSAT,theideathat make deisions to satisfy reently dedued lauses[4, 5] has een proven to e very effetive for solving hard ominational prolems. However, when applying a ominational solver for sequential SAT, the run time of ominational SAT solving is not neessarily the most important fator to e optimized. In this paper, we propose a deision variale seletion heuristi whih is more suitale for solving sequential prolems. The rest of the paper is organized as the following. Setion II illustrates the use of ominational SAT in sequential SAT. In Setion III, we disuss the importane of state redution and desrie our approahes to this prolem. Setion IV presents our sequential searh framework. In Setion V, we present our priority-ased searh strategy. Setion VI disusses the impat of deision ordering in ominational SAT on the overall sequential SAT effiieny. Setion VII summarizes experimental results to demonstrate the superiority of our urrent sequential SAT solver. Setion VIII onludes the paper.

2 II. Cominational and sequential SAT Given a finite set of variales, V, over the set of Boolean values B {0,1}, aliteral, l/l is an instane of a variale or its omplement, v/ v V. A lause i, is a disjuntion of literals (l 1 l 2... l n ). Aformula f, is a onjuntion of lauses n. A lause is onsidered as a set of literals, and a formula as a set of lauses. An assignment A satisfies a formula f if f (A)=1. An assignment is alled maximal when every variale in V reeives a value assignment. Given a sequential iruit, a state lause is a lause onsisting only of state variales. In a Boolean Satisfiaility (SAT) prolem (or here all it ominational SAT to differentiate from sequential SAT), a formula f is given and the prolem is to find an assignment A to satisfy f or prove that no suh an assignment exists. SAT has attrated tremendous researh effort in reent years, resulting in the developments of various effiient SAT solver pakages suh as [4, 5, 6, 10, 11]. Through akward timeframe expansion (Figure 1), a sequential SAT prolem an e translated into a sequene of ominational SAT prolems. II-A. Sequential SAT and state lauses Initial state: (x=1, y=0, z=1) (1) time frame n time frame n a 1 a (2) f=1 0 0 f=1 PPI x PPI y PPI z PPO x 0 PPI x 1 PPI y 0 PPI z PPO x time frame (n 1) state time frame n a ojetive a f=1 PPI x PPO x 0 PPI x PPO x (3) PPI y PPI z 1 PPI y PPI z state lause: (x+y ) state lause: (x+y )(x) (4) No solution: UNSAT Figure 2: Sequential SAT with state lauses To illustrate the usage of state lauses in sequential SAT, Figure 2 depits a simple example iruit with three primary inputs a,,, one primary output f, and three state elements x,y,z. The initial state is (x = 1,y = 0,z = 1). Suppose the SAT ojetive is to satisfy f = 1. Starting from time-frame n where n is unknown, the iruit is translated into a ominational opy with state variales dupliated as PPOs and PPIs. This is illustrated as (1) in the Figure. Sine this represents a ominational SAT prolem, a Boolean SAT solver an e applied. Suppose after the ominational SAT solving, a solution (a = 1, = 1, = 0,PPI x = 0,PPI y = 1,PPI z = 0) for satisfying f = 1 (step (2))is identified. Notie that all inputs have values assigned in this solution, even though it might not e neessary to have all of them assigned in order to satisfy f = 1. This phenomenon is due to the use of a ominational SAT solver as mentioned earlier. The assignments at PPIs implies a solution state (x = 0,y = 1,z = 0) that doesn t ontain the initial state. Instead of justifying the state (x = 0,y = 1,z = 0) in time-frame n 1, we an examine it first to hek whether we an hange as many of the assignments to don t-ares while still satisfying the ojetive f = 1. Suppose after analysis, we determine that z = 0 is unneessary, it an e removed from the solution state (x = 0,y = 1,z = 0) and a solution state (x = 0,y = 1) is derived. We all this step solution state redution sine it tries to remove the unneessary state variale assignments from a solution. The solution state (x = 0,y = 1) doesn t ontain the initial state either, so it still needs to e justified in timeframe n 1. Bakward time-frame expansion an e ahieved y adding a state ojetive (PPO x = 0, = 1) to the ominational opy of the iruit. Also, a state lause (x+y ) is generated to prevent reahing state solutions ontained y the state su-spae (x = 0,y = 1) in time-frame n 1. The new ominational SAT instane is then passed to the Boolean SAT solver. Suppose in time-frame n 1, no solution an e found for state ojetive (PPO x = 0, = 1). Then, we aktrak to time-frame n to find another solution. In a way, we have proved that from state (x = 0,y = 1), there exists no solution. Therefore, there is no need to remove the state lause (x + y ). However, at this point it is enefiial to perform further analysis to determine if oth PPO x = 0and = 1 are involved in the onflit that indiates state ojetive (PPO x = 0, = 1) is unsatisfiale. Suppose after onflit analysis, It is disovered that only PPO x = 0 is involved in the onflit. A state lause (x) is then added. The future ominational SAT instanes will then have the state lauses (x + y )(x) inluded, that reord the non-solution state su-spaes previously identified. This is illustrated in (4) of the Figure. We all this step ojetive state redution sine it tries to derive a smaller state lause when a state ojetive eomes unsatisfiale. The solving ontinues until either one of the following two onditions is reahed: 1. After solution state redution, a solution is found whose state part ontains the initial state. For example, a solution with the state part (x = 1,z = 1) ontains the initial state (x = 1,y = 0,z = 1). In this ase, a solution for the sequential SAT prolem is found. Note that a solution without any assignments to PPIs ontains any initial state. 2. If aktraked to time-frame n, the initial ojetive f = 1 annot e satisfied under the onstraints imposed y all the state lauses added, then the original prolem is unsatisfiale. This is equivalent to say: if any ojetive (inluding the initial ojetive f = 1 and all intermediate ojetives) annot e satisfied under the onstraints imposed y all the state lauses added, then the original prolem is unsatisfiale. The aove example illustrates several important onepts in the design of our urrent sequential SAT solver. The state redution involves finding smaller state lauses in order to more effetively prune the searh spae. There are two types of state redution. One is solution state redution, the other is ojetive state redution. The use of state lauses serves two purposes: (1) to reord those state su-spaes that have een explored, and (2) to reord those state su-spaes ontaining no solution. The first is to prevent the searh from entering a state justifiation loop. Although oneptually the searh follows the akward timeframe expansion, the aove example demonstrates that in the implementation, expliit time-frame expansion is not neessary. In other words, a sequential SAT solver needs only one opy of the iruit. Moreover, there is no need to memorize the numer of time-frames eing expanded. Later we will disuss our implementation of the solver and demonstrate how this an e ahieved y using an ojetive list. The aove example demonstrates the use of depth-first searh strategy. However, y using state lauses, more flexile searh strategies are feasile if the intermediate state ojetives are reorded in a list. The idea of the sequential searh framework an e desried as follows: Given a sequential prolem with an ojetive oj and an initial state s 0, C is the one time-frame ominational opy of the sequential iruit where eah state variale is expanded to a PPI variale and a PPO variale. Suppose we use an ojetive list FO to store the intermediate ojetives, and initially it only ontains oj. Eah time, an ojetive o is seleted from FO and given to the ominational iruit solver.

3 The iruit solver then solves o on C. There will e three possile outomes reported y the solver: (1) o is unsatisfiale, (2) at least one state in the otained solutions ontains the initial state s 0, and (3) no state in the otained solutions ontains the initial state s 0. For ase (1), we an remove o from FO. For ase (2), the searh stops and the sequential prolem is proven satisfiale. For ase (3), the state part of eah solution is added to FO as a new state ojetive and a state lause is added to C to prevent the state e explored and found again in any future solutions. If finally, the ojetive list FO eomes empty, whih means all ojetives have een proven to e unsatisfiale, then the searh stops and the sequential prolem is proven unsatisfiale. In this framework, the keys are to deide how to selet the next ojetive from the list for solving and to deide how many solutions to e found for an ojetive eah time. The details will e giveninsetioniv. In the following, we will desrie the detail of Seq-SAT. To failitate the desription, we define the following two terms: (1) a frame ojetive is an ojetive to e satisfied whih is passed to the ominational SAT solver. A frame ojetive an e either the initial ojetive or a state ojetive. (2) a frame solution is an assignment at the PIs and PPIs, whih satisfies a given frame ojetive. III. State redution for state spae pruning There are two types of state redution: solution state redution (SSR) and ojetive state redution (OSR). SSR is performed when a frame solution is identified y the ominational SAT solver. OSR is applied when a state ojetive eomes unsatisfiale. The details are disussed in the following. III-A. Solution state redution When a frame solution is identified y the ominational SAT solver, all PIs and PPIs have assigned values. Our goal is to find and un-assign the unneessary assignments at the PPIs. With fewer assigned PPIs, a smaller state lause an e reated, whih prunes a larger state su-spae. Given an initial frame solution, to derive one ontaining a minimal numer of state variales with value assignment is an NP-omplete prolem. Therefore, we only derive a loally minimized 3-value solution. A frame solution s of a frame ojetive o is a loally minimized 3-value solution if and only if un-assign any state variale of s (hanging its value from a 0/1 to X), the resulting assignments annot satisfy o y 3-value simulation. One straight forward heuristi to derive a loally minimized 3- value solution is to employ 3-value simulation. That is, for all assigned state variales of a solution, un-assign one of them at a time to hek if its assignment is neessary to satisfy the frame ojetive. If not, the assignment is removed from the solution, this proedure ontinues until eah assigned state variale has een heked one, then the remaining assignments form a 3-value minimized solution. A frame solution reported y a SAT solver usually ontains many unneessary assignments at state variales. Diretly apply 3-value simulation ould e omputationally expensive. For example, removing N unneessary state assignments needs at least N times of simulation even if parallel simulation is used. To redue the numer of simulation runs and improve effiieny, we adopt a two-step state redution approah. In this approah, a modified D-algorithm [8] is first applied to otain a minimized solution. Given a frame ojetive O and its frame solution S, the modified D-algorithm justify O ased on S, that is, when the D-algorithm makes a deision at a signal, the value assigned to the signal has to e onsistent with its value in the solution S. Hene, the D-algorithm is used as a trae proedure, not as a searh proedure. Based on the minimized solution from the first step, 3-value simulation is then employed to derive a 3-value minimized solution. This approah is ased on the assumption that the D-algorithm an usually find a solution ontaining muh fewer assignments than that of a SAT solver. Experiments show our assumption is valid in general. The details of the approah are omitted due to page limit. TABLE I: RESULTS OF SOLUTION STATE REDUCTION Ciruit Org # Step 1 resulting Step 2 resulting # %redued times(se) # %redued times(se) s % % 1 s % % 1 s % % 2 s % % 3 s % % 100 s % % 110 Tale I shows experimental results to demonstrate the impat of solution state redution. All our experiments were run on P4 2GHz Linux mahines with 1 GB memory. In these experiments, a sequene of sequential SAT prolem instanes are reated for eah iruit. For eah iruit and eah of its primary outputs f, the first SAT ojetive is to satisfy f = 1 and then we would try to satisfy f = 0. This is a typial primary output toggle experiment. The results shown for eah iruit are for solving all toggle SAT instanes for the iruit. We assume the initial state is that all state elements have a 0 value. Note, this kind of prolems may ontain very diffiult ases even for ISCAS 89 iruits. The two # olumns show the numers after state redution with modified D-algorithm (step 1) and then, with three value simulation (step 2), respetively. The two % redued olumns show the orresponding perentage of redution. The two run-time olumns show the run times ased on using step 1 and then, using step 2. We note that if no state redution is performed, then the sequential SAT might not finish solving all instanes within a reasonale time for eah of these examples. From the results, we an see that the modified D-algorithm ould redue most of the unneessary assignments to the state variales, so the numer of 3-value simulation runs an e greatly redued. Although step 2 does not help muh in terms of the total perentage of assignment redution, it does help further improve the overall performane. III-B. Ojetive state redution As a state so is added to the ojetive list, a state lause s is added to prevent the state from eing reahed again. When the state ojetive so eomes unsatisfiale, instead of using the same state lause s, a smaller state lause might e added to prune a larger searh spae. This an e ahieved y additional onflit analysis when a state ojetive is proved unsatisfiale. We explain the idea y an example, suppose the state ojetive so is (PPO r = 1,PPO s = 1,PPO t = 0,PPO x = 0, = 1), the impliation graph when s is proved unsatisfiale (whih means there is a onflit that an not e resolved y aktrak) is shown as in Figure 3. From Figure 3, if traing from the onflit points to the PPO state variales, we an dedue that (PPO r = 1,PPO s = 1,PPO t = 0) is unsatisfiale, so the state lause (r + s + t), whih is smaller than (r + s + t + x + y ), an e added to the iruit. However, we need to e areful that the state lauses added in this way shouldn t exlude the searh spae that ontains the initial state. Otherwise a satisfiale prolem erroneously eomes unsatisfiale sine the initial state ouldn t e reahed anymore after adding suh a state lause. Using the aove example, if the values of r, s and t in the initial state are 1, 1 and 0 respetively, state lause (r + s +t) would exlude the initial state. To avoid this, we an just selet a state variale from the state ojetive whose ojetive value is not its initial value and add its orresponding literal to the state lause. For example, if the value of x in the initial state is

4 1 in the aove example, we an add x to the state lause (r + s +t) to form state lause (r + s + t + x ) whih an e added to the iruit. This an always e done sine state ojetives don t ontain the initial state. That is, eah state ojetive must have at least one variale whose ojetive value is not same as its value in the initial state, otherwise, the prolem should has een proved satisfiale. This redution is valid even if state lauses that were used for preventing justifiation loop are involved in the onflit graph. We also note that this redution is not related to identify the suset of a CNF formula that is suffiient for unsatisfiaility, instead, it just tries to find a onflit lause that ontains only literals of the variales in a state ojetive. Tale II gives the experimental results omparing the performane of our sequential solver with and without OSR. These are the same toggle experiments as those in Tale I. TABLE II: EFFECTS OF OBJECTIVE STATE REDUCTION Without OSR With OSR Ciruit runtimes(se) stats runtimes(se) stats s /0/0 1 12/0/0 s /0/0 1 10/0/0 s /5/4 3 90/8/0 s /58/ /63/13 s /54/ /69/5 s /0/ /0/0 s /0/ /0/0 s /58/ /58/0 **aort time for eah prolem is 100 se. stats: x/y/z are #SAT/#UNSAT/#Aort From the experiment results, it an e seen that the performane is greatly improved for half of the testases y OSR, and only for one testase, it auses a small overhead. PPO PPO t PPO x PPO s y PPO r a e d f f Confliting Variale Figure 3: An impliation graph when a state ojetive is proved unsatisfi ale IV. Flexile sequential searh framework In our sequential SAT design, state lauses are aumulated through the solving proess. The sequential solving proess onsists of a sequene of ominational solving tasks ased on a given frame ojetive. At the eginning, the frame ojetive is the initial ojetive. As the solving proeeds, many frame solutions eome frame ojetives. These frame ojetives are stored in the ojetive list. The use of state lauses and the ojetive list offer muh flexiility in the hoie of the searh strategy. Different searh strategies an e implemented y using different riteria in seleting the next frame ojetive for solving, and in deiding how many solutions to e otained at a time for a frame ojetive. For example, for the depth-first searh, the ojetive list is used as a stak where the next seleted ojetive is the one most reently produed. For the readth-first searh, the ojetive list is used as a queue. In addition, in the readth-first searh, it demands the solver to find all solutions for the seleted ojetive. After that, the ojetive is removed from the ojetive list. This framework further enales us to explore various searh strategies etween these two extreme ones whih will e desried later in this setion. The overall framework of our sequential SAT solver is desried in Algorithm IV.1. In the desription, we assume that eah frame ojetive produes at most one frame solution at a time. This assumption is only for purpose of easier explanation. The algorithm an e easily modified to serve the ase that multiple solutions are produed at a time. The proedure PPO state onflit analysis() does ojetive state redution, and state redution() does solution state redution. The proedure selet a frame ojetive() selets a frame ojetive to e solved next, upon different implementation, various searh strategies an e realized, and the details will e given in the next setion. A frame ojetive an e removed from the ojetive list only if it is proven unsatisfiale y the ominational SAT. If it is satisfiale, the frame ojetive stays in the ojetive list. If the ojetive list eomes empty during the sequential searh, it means that the solver has exhausted all the state ojetives and, therefore, proved that the initial ojetive is unsatisfiale. Note that during eah step of ominational SAT, onflit lauses are also aumulated through the ominational SAT solving proess. When the sequential solving swithes from one frame ojetive to another, these onflit lauses stay. Hene, in the sequential solving proess, the onflit lauses generated y the ominational SAT are also aumulated. Our experiene indiates that although these onflit lauses does help speed up the ominational SAT solving, for sequential SAT, the state lauses are the dominating soure for the effiieny improvement of the overall sequential searh. Algorithm IV.1: SEQUENTIALSOLVER(C,o j,s 0 ) omment: C is the iruit with PPIs and PPOs expanded omment: o j is the initial ojetive omment: s 0 is the initial state omment: FO is the ojetive list FO {o j}; while (FO /0) f o j selet a frame ojetive(fo); f sol ominational solve a frame ojetive(c, f o j ); if ( f sol = NULL) lause PPO state onflit analysis(c, f o j ); add state lause(c,lause); FO FO {f o j }; do state ue state redution(c, f o j, f sol ); if { (s 0 state ue) return else (SAT); lause onvert to lause(state ue); else add state lause(c,lause); FO FO+ {state ue}; return (UNSAT) V. On the sheduling of frame ojetives In this setion, we introdue a new searh strategy, alled the priority-ased searh strategy, whih an e easily implemented in the searh framework desried aove. We further ompare it with the depth-first and readth-first searh strategies and demonstrate its superiority. Given a state ojetive, we define its size as the total numer of state variales minus the numer of state variales with an inary ojetive value (i.e. not an unknown X) in the state ojetive. We use size(o j) to denote the size of o j. Note that a state ojetive with alargersize orresponds a larger state su-spae. Hene, if there are multiple state ojetives in the ojetive list, it seems intuitive

5 to hoose the largest state ojetive as the next ojetive for solving as it orresponds to exploring the solution with the largest state su-spae, and its solutions, if any, tend to ontain fewer state assignments, whih means smaller state lauses are likely to e generated to prune a larger searh spae. We further define the ount of an ojetive o j as the numer of times that o j has een seleted for solving y the ominational solver. If we use ount to guide the seletion of ojetives and hoose the one with the largest ount as the next ojetive, it will e more like the readth-first searh strategy. On the other hands, if we hoose the one with the smallest ount, It will e more like the depth-first searh. Now we define priority of a frame ojetive o j, pri(o j), as follows: pri(o j) =size(o j) ount(o j)/10 (1) In our implementation of the priority-ased searh strategy, we selet the ojetive with the largest value of pri. If there is a tie under this seletion riterion, then the most reently generated one will e seleted. For eah seleted ojetive, at most one solution is otained at a time. Through extensive experiments, we oserve that omining size and ount in this way maximizing the performane, in general, for oth satisfiale and unsatisfiale ases. With the ojetive list, depth-first and readth-first searh strategies an e implemented easily. In depth-first searh, the list ehaves like a stak. Hene, the seleted ojetive is the one most reently produed. In readth-first searh, the list ehaves like a queue. However, in readth-first searh, solving an ojetive means to finish produing all possile solutions. Then, the ojetive is removed from the ojetive list. Tale III gives the experimental results omparing different searh strategies. These are the same toggle experiments as those in Tale I. The priority olumn is the same as the last times olumn in Tale I. As it an e seen, the new searh strategy signifiantly outperforms the depth-first and the readth-first strategies. TABLE III: CPU RUNTIMES (IN SECONDS) ON SEARCH STRATEGIES depth-fi rst readth-fi rst priority Ciruit runtimes stats runtimes stats runtimes stats (se) (se) (se) s /0/0 2 12/0/0 1 12/0/0 s /0/ /0/1 1 10/0/0 s /8/ /5/4 3 90/8/0 s /54/ /63/ /63/13 s /60/ /59/ /69/5 s /0/ /0/ /0/0 s /0/ /0/ /0/0 s /57/ /57/ /58/0 **aort time for eah prolem is 100 se. stats: x/y/z are #SAT/#UNSAT/#Aort VI. On the deision variale seletion heuristis in the ominational SAT solver For pure ominational SAT, oth VISDS[4] deision variale seletion and lause ased deision variale seletion [5] have een proven effetive. However, when applying a ominational SAT solver for sequential SAT, the runtime of ominational SAT solving is not the only ritial fator. In eah run of ominational SAT solving during the sequential SAT, the smaller the neessary assignments are made at the PPIs, the more effiiently the state spae an e pruned. Therefore, the deision variale seletion heuristis in ominational SAT also need modifiations. To demonstrate this point, we onduted experiments to ompare different deision variale seletion heuristis in ominational SAT. Deision heuristi 1: This is the same as the VISDS deision variale seletion [4]. Deision heuristi 2: First, the deision variale is seleted from the most reently generated onflit lause that is not yet satisfied [5]. If all the generated onflit lauses are satisfied, the deision variale is seleted ased on the J-nodes [17, 6] with the highest topologial order (losest to the POs and the PPOs). This deision heuristi is used in the original ominational SAT solver C-SAT [6]. Deision heuristi 3: First, the deision variale is seleted from the most reently generated onflit lause that is not yet satisfied as in Heuristi 2 aove. If all the generated onflit lauses are satisfied, the deision variale is seleted from PIs and PPIs where PIs are seleted efore PPIs This is the default heuristi used in our final implementation. The reasons we adopt Heuristi 3 in Seq-SAT are summarized as follows: (1). Using input variales as deision points results in etter performane for easy ominational prolems in general. For a sequential prolem that an e transformed into a series of relatively easy ominational prolems, this heuristi works very well. (2). For hard prolems, as the onflit lauses tend to aumulate fast during searh, the deisions will gradually e dominated y the onflit lauses. Therefore, the performane will not degrade muh. (3). Seleting PIs prior to PPIs as deision variales tends to result in solutions with fewer state variales with value assignments [14]. TABLE IV: CPU RUNTIMES ON DIFFERENT DECISION HEURISTICS heuristi 1 heuristi 2 heuristi 3 Ciruit runtimes stats runtimes stats runtimes stats (se) (se) (se) s /0/0 3 12/0/0 1 12/0/0 s /0/2 1 10/0/0 1 10/0/0 s /7/ /7/1 3 90/8/0 s /62/ /63/ /63/13 s /69/ /69/ /69/5 s /0/ /0/ /0/0 s /0/ /0/ /0/0 s /58/ /58/ /58/ /1/0 13 0/1/0 20 0/1/ /1/0 16 0/1/0 8 0/1/0 **aort time for eah prolem is 100 se. stats: x/y/z are #SAT/#UNSAT/#Aort Tale IV shows the results of different heuristis. The experiments for s- iruits were the same PO toggle experiments iruit as explained efore. The experiments for - iruits were the ominational equivalene heking prolem using the miter iruit of two idential ISCAS 85 iruits. It an e seen that the deision heuristi 3 works well for sequential iruits, even for large ominational iruits, its performane is not ad. VII. Additional experimental results Tale V ompares our solver with NuSMV [16] ounded model heking funtion for satisfiale ases. We have shown in the previous experiments how eah idea works, Tale VI ompares the overall performane of Seq-SAT with Satori. In Tale V, the Property olumn gives the LTL formula for eah testase, and length olumns give the length of the witness vetors otained y Seq-SAT and NuSMV respetively. For NuSMV, the version we used is NuSMV zhaff in whih the SAT solver zhaff is used for ounded model heking. The initial states used in these examples are all 0.

6 TABLE V: COMPARISON TO NUSMV BOUNDED MODEL CHECKING Seq-SAT NuSMV Ciruit Property runtimes(se) length runtimes(se) length s526 G(g214 = 0) s13207 G(g594 = 0) s13207 G(g785 = 0) s38417 G(g5549 = 0) > s38417 G(g16399 = 0) s38584 G(g11678 = 0) s38584 G(g29212 = 0) > **aort time for eah prolem is 3600s. From Tale V, it an e seen that, even though the witness vetors derived y Seq-SAT are in general longer than those y NuSMV, Seq- SAT learly outperforms NuSMV in terms of CPU runtime. In Satori, the depth-first searh and the readth-first searh are implemented separately. So in Tale VI, we ompare Seq-SAT with Satori under these two searh strategies for the same toggle experiments as those in Tale I. Tale VII summarizes the numer of the aort ases shown in Tale VI. Column #Satori aort shows the numer of ases aorted in oth Satori depth-first searh and Satori readth-first searh. Column #SSANA gives the numer of ases aorted in Seq-SAT ut not aorted in either Satori depth-first searh or Satori readth-first searh. TABLE VI: COMPARISON TO SATORI Satori depth-fi rst Satori readth-fi rst Seq-SAT Ciruit runtimes stats runtimes stats runtimes stats (se) (se) (se) s /0/4 5 12/0/0 1 12/0/0 s /0/ /0/1 1 10/0/0 s /4/ /4/ /8/0 s /48/ /59/ /63/13 s /52/ /52/ /69/5 s /0/ /0/ /0/0 s /0/ /0/ /0/0 s /57/ /58/ /58/0 **aort time for eah prolem is 100 se. stats: x/y/z are #SAT/#UNSAT/#Aort TABLE VII: ABORTED CASES ANALYSIS Ciruit #Satori aort #Seq-SAT aort #SSANA s s s s s s s s It an e seen that Seq-SAT ahieves signifiant performane improvement for almost all ases. It is also interesting to note that the results of the depth-first searh in Satori are quite different from those of depth-first searh implemented in Seq-SAT (see Setion V Tale III). This indiates that the depth-first searh is quite unstale (espeially for satisfiale ases) and easy falls into ad searh area. VIII. Conlusion In this paper, we present an effiient sequential SAT solver Seq- SAT. The implementation of Seq-SAT employs four new ideas: (1) etter state redution algorithms, (2) a more flexile searh framework for aommodating and integrating different searh strategies, (3) the priority-ased searh strategy, and (4) a modified deision variale seletion heuristi for the underlying ominational iruit SAT solver. With these new ideas and effiient data strutures and implementation, Seq-SAT ahieves very signifiant speedup over Satori. We will release the soure ode of Seq-SAT along with the puliation of this paper. Referenes [1] M. K. Iyer, G Parthasarathy, K.-T. Cheng, SATORI - A Fast Sequential SAT Engine for Ciruits. Pro. IEEE/ACM International Conferene on Computer-Aided Design [2] A. Biere, A. Cimatti, E. M. Clarke, M. Fujita, and Y. Zhu, Symoli Model Cheking using SAT Proedures instead of BDDs. In Pro. of the 36th ACM/IEEE Design Automation Conf., pp , June [3] P. Tafertshofer, A. Ganz, and K. Antreih, IGRAINE An Impliation-Graph ased Engine for Fast Impliation, Justifiation and Propagation. IEEE Trans. on Computer-Aided Design, 19(8): , August [4] M. Moskewiz, C. Madigan, Y. Zhao, L. Zhang, and S. Malik, Chaff: Engineering an effiient SAT solver. Pro. of ACM/IEEE Design Automation Conf., pp , June [5] E. Golderg and Y. Novikov, BerkMin: A fast and roust Sat Solver. Pro. European Design and Test Conf., pp , Marh [6] Feng Lu, Li-C. Wang, Kwang-Ting Cheng, A iruit SAT solver with signal orrelation guided learning. In Pro. European Design and Test Conferene [7] P. Tafertshofer, A. Ganz, and M. Henftling, A SAT-Based Impliation Engine for Effiient ATPG, Equivalene Cheking, and Optimization of Netlists. Pro. IEEE/ACM International Conferene on Computer-Aided Design, pp , Nov [8] M. Aramovii, M. A. Breuer, and A. D. Friedman. Digital Systems Testing and Testale Design. CS Press, 1 st ed., [9] G Parthasarathy, M. K. Iyer, Li-C. Wang and K.-T. Cheng, Effiient Reahaility Cheking and Pre-image Computation using Sequential SAT. Pro. Asia and South Paifi Design Automation Conf [10] H. Zhang. SATO: An Effiient Propositional Prover. Pro. of International Conferene on Automated Dedution, Vol 1249, LNAI, pp , 1997 [11] J. P. Marques-Silva and K. A. Sakallah. GRASP: A Searh Algorithm for Propositional Satisfiaility. IEEE Trans on Computers, vol.48, pp , [12] L. Zhang, C. Madigan, M. Moskewiz, and S. Malik. Effiient onflit driven learning in a Boolean satisfiaility solver. Pro. IEEE/ACM International Conferene on Computer-Aided Design, pp , Nov [13] K. L. MMillan. Applying SAT methods in unounded symoli model heking. Pro. of International Conferene on Computer Aided Verifiation, Vol 2404, LNSC, pp , [14] H.-J. Kang, I.-C. Park. SAT-Based Unounded Symoli Model Cheking. Pro. of ACM/IEEE Design Automation Conf., pp , June [15] B. Li, M. S. Hsiao, and S. Sheng. A novel SAT all-solutions solver for effiient preimage omputation. Pro. of Design, Automation and Test in Europe Conf., pp , Fe [16] A. Cimatti, E. Clarke, E. Giunhiglia, et. al.. NuSMV 2: An OpenSoure Tool for Symoli Model Cheking. In Pro. CAV 02, LNCS. Springer Verlag, [17] M.K.Ganai, L.Zhang, P.Ashar, A.Gupta, and S.Malik. Comining strengths of iruit-ased and CNF-ased algorithms for a high-performane SAT solver. In Pro. ACM/IEEE Design Automation Conf., pp , June 2002.

Formal Verification SE303 Conception des systèmes sur puces (SoC)

Formal Verification SE303 Conception des systèmes sur puces (SoC) Formal Verifiation SE33 Coneption des systèmes sur pues (SoC) Ulrih Kühne 29//27 Outline Introdution Short History of Hardware Failures Design and Verifiation Proess Funtional Verifiation Ciruit Models

More information

Sufficient Conditions for a Flexible Manufacturing System to be Deadlocked

Sufficient Conditions for a Flexible Manufacturing System to be Deadlocked Paper 0, INT 0 Suffiient Conditions for a Flexile Manufaturing System to e Deadloked Paul E Deering, PhD Department of Engineering Tehnology and Management Ohio University deering@ohioedu Astrat In reent

More information

I F I G R e s e a r c h R e p o r t. Minimal and Hyper-Minimal Biautomata. IFIG Research Report 1401 March Institut für Informatik

I F I G R e s e a r c h R e p o r t. Minimal and Hyper-Minimal Biautomata. IFIG Research Report 1401 March Institut für Informatik I F I G R e s e a r h R e p o r t Institut für Informatik Minimal and Hyper-Minimal Biautomata Markus Holzer Seastian Jakoi IFIG Researh Report 1401 Marh 2014 Institut für Informatik JLU Gießen Arndtstraße

More information

SRC Research. Report. An Efficient Matching Algorithm for a High-Throughput, Low-Latency Data Switch. Thomas L. Rodeheffer and James B.

SRC Research. Report. An Efficient Matching Algorithm for a High-Throughput, Low-Latency Data Switch. Thomas L. Rodeheffer and James B. Novemer 5, 998 SRC Researh Report 6 An Effiient Mathing Algorithm for a High-Throughput, Low-Lateny Data Swith Thomas L Rodeheffer and James B Saxe Systems Researh Center 30 Lytton Avenue Palo Alto, CA

More information

EE 321 Project Spring 2018

EE 321 Project Spring 2018 EE 21 Projet Spring 2018 This ourse projet is intended to be an individual effort projet. The student is required to omplete the work individually, without help from anyone else. (The student may, however,

More information

A population of 50 flies is expected to double every week, leading to a function of the x

A population of 50 flies is expected to double every week, leading to a function of the x 4 Setion 4.3 Logarithmi Funtions population of 50 flies is epeted to doule every week, leading to a funtion of the form f ( ) 50(), where represents the numer of weeks that have passed. When will this

More information

Lightpath routing for maximum reliability in optical mesh networks

Lightpath routing for maximum reliability in optical mesh networks Vol. 7, No. 5 / May 2008 / JOURNAL OF OPTICAL NETWORKING 449 Lightpath routing for maximum reliability in optial mesh networks Shengli Yuan, 1, * Saket Varma, 2 and Jason P. Jue 2 1 Department of Computer

More information

Counting Distinct Objects over Sliding Windows

Counting Distinct Objects over Sliding Windows Counting Distint Ojets over Sliding Windows Wenjie Zhang, Ying Zhang Muhammad Aamir Cheema Xuemin Lin, the University of New South Wales {zhangw,yingz,maheema,lue}@se.unsw.edu.au NICTA Astrat Aggregation

More information

Nonreversibility of Multiple Unicast Networks

Nonreversibility of Multiple Unicast Networks Nonreversibility of Multiple Uniast Networks Randall Dougherty and Kenneth Zeger September 27, 2005 Abstrat We prove that for any finite direted ayli network, there exists a orresponding multiple uniast

More information

Math 151 Introduction to Eigenvectors

Math 151 Introduction to Eigenvectors Math 151 Introdution to Eigenvetors The motivating example we used to desrie matrixes was landsape hange and vegetation suession. We hose the simple example of Bare Soil (B), eing replaed y Grasses (G)

More information

The Complete Energy Translations in the Detailed. Decay Process of Baryonic Sub-Atomic Particles. P.G.Bass.

The Complete Energy Translations in the Detailed. Decay Process of Baryonic Sub-Atomic Particles. P.G.Bass. The Complete Energy Translations in the Detailed Deay Proess of Baryoni Su-Atomi Partiles. [4] P.G.Bass. PGBass P12 Version 1..3 www.relativitydomains.om August 218 Astrat. This is the final paper on the

More information

Taste for variety and optimum product diversity in an open economy

Taste for variety and optimum product diversity in an open economy Taste for variety and optimum produt diversity in an open eonomy Javier Coto-Martínez City University Paul Levine University of Surrey Otober 0, 005 María D.C. Garía-Alonso University of Kent Abstrat We

More information

The Effectiveness of the Linear Hull Effect

The Effectiveness of the Linear Hull Effect The Effetiveness of the Linear Hull Effet S. Murphy Tehnial Report RHUL MA 009 9 6 Otober 009 Department of Mathematis Royal Holloway, University of London Egham, Surrey TW0 0EX, England http://www.rhul.a.uk/mathematis/tehreports

More information

FORMAL METHODS LECTURE VI BINARY DECISION DIAGRAMS (BDD S)

FORMAL METHODS LECTURE VI BINARY DECISION DIAGRAMS (BDD S) Alessandro Artale (FM First Semester 2009/2010) p. 1/38 FORMAL METHODS LECTURE VI BINARY DECISION DIAGRAMS (BDD S) Alessandro Artale Faulty of Computer Siene Free University of Bolzano artale@inf.unibz.it

More information

Reliability Guaranteed Energy-Aware Frame-Based Task Set Execution Strategy for Hard Real-Time Systems

Reliability Guaranteed Energy-Aware Frame-Based Task Set Execution Strategy for Hard Real-Time Systems Reliability Guaranteed Energy-Aware Frame-Based ask Set Exeution Strategy for Hard Real-ime Systems Zheng Li a, Li Wang a, Shuhui Li a, Shangping Ren a, Gang Quan b a Illinois Institute of ehnology, Chiago,

More information

Searching All Approximate Covers and Their Distance using Finite Automata

Searching All Approximate Covers and Their Distance using Finite Automata Searhing All Approximate Covers and Their Distane using Finite Automata Ondřej Guth, Bořivoj Melihar, and Miroslav Balík České vysoké učení tehniké v Praze, Praha, CZ, {gutho1,melihar,alikm}@fel.vut.z

More information

7 Max-Flow Problems. Business Computing and Operations Research 608

7 Max-Flow Problems. Business Computing and Operations Research 608 7 Max-Flow Problems Business Computing and Operations Researh 68 7. Max-Flow Problems In what follows, we onsider a somewhat modified problem onstellation Instead of osts of transmission, vetor now indiates

More information

Millennium Relativity Acceleration Composition. The Relativistic Relationship between Acceleration and Uniform Motion

Millennium Relativity Acceleration Composition. The Relativistic Relationship between Acceleration and Uniform Motion Millennium Relativity Aeleration Composition he Relativisti Relationship between Aeleration and niform Motion Copyright 003 Joseph A. Rybzyk Abstrat he relativisti priniples developed throughout the six

More information

Modeling of discrete/continuous optimization problems: characterization and formulation of disjunctions and their relaxations

Modeling of discrete/continuous optimization problems: characterization and formulation of disjunctions and their relaxations Computers and Chemial Engineering (00) 4/448 www.elsevier.om/loate/omphemeng Modeling of disrete/ontinuous optimization problems: haraterization and formulation of disjuntions and their relaxations Aldo

More information

Parallel disrete-event simulation is an attempt to speed-up the simulation proess through the use of multiple proessors. In some sense parallel disret

Parallel disrete-event simulation is an attempt to speed-up the simulation proess through the use of multiple proessors. In some sense parallel disret Exploiting intra-objet dependenies in parallel simulation Franeso Quaglia a;1 Roberto Baldoni a;2 a Dipartimento di Informatia e Sistemistia Universita \La Sapienza" Via Salaria 113, 198 Roma, Italy Abstrat

More information

Variation Based Online Travel Time Prediction Using Clustered Neural Networks

Variation Based Online Travel Time Prediction Using Clustered Neural Networks Variation Based Online Travel Time Predition Using lustered Neural Networks Jie Yu, Gang-Len hang, H.W. Ho and Yue Liu Abstrat-This paper proposes a variation-based online travel time predition approah

More information

Ordering Generalized Trapezoidal Fuzzy Numbers. Using Orthocentre of Centroids

Ordering Generalized Trapezoidal Fuzzy Numbers. Using Orthocentre of Centroids International Journal of lgera Vol. 6 no. 69-85 Ordering Generalized Trapezoidal Fuzzy Numers Using Orthoentre of Centroids Y. L. P. Thorani P. Phani Bushan ao and N. avi Shanar Dept. of pplied Mathematis

More information

Complete Shrimp Game Solution

Complete Shrimp Game Solution Complete Shrimp Game Solution Florian Ederer Feruary 7, 207 The inverse demand urve is given y P (Q a ; Q ; Q ) = 00 0:5 (Q a + Q + Q ) The pro t funtion for rm i = fa; ; g is i (Q a ; Q ; Q ) = Q i [P

More information

IMPEDANCE EFFECTS OF LEFT TURNERS FROM THE MAJOR STREET AT A TWSC INTERSECTION

IMPEDANCE EFFECTS OF LEFT TURNERS FROM THE MAJOR STREET AT A TWSC INTERSECTION 09-1289 Citation: Brilon, W. (2009): Impedane Effets of Left Turners from the Major Street at A TWSC Intersetion. Transportation Researh Reord Nr. 2130, pp. 2-8 IMPEDANCE EFFECTS OF LEFT TURNERS FROM THE

More information

DIGITAL DISTANCE RELAYING SCHEME FOR PARALLEL TRANSMISSION LINES DURING INTER-CIRCUIT FAULTS

DIGITAL DISTANCE RELAYING SCHEME FOR PARALLEL TRANSMISSION LINES DURING INTER-CIRCUIT FAULTS CHAPTER 4 DIGITAL DISTANCE RELAYING SCHEME FOR PARALLEL TRANSMISSION LINES DURING INTER-CIRCUIT FAULTS 4.1 INTRODUCTION Around the world, environmental and ost onsiousness are foring utilities to install

More information

A Spatiotemporal Approach to Passive Sound Source Localization

A Spatiotemporal Approach to Passive Sound Source Localization A Spatiotemporal Approah Passive Sound Soure Loalization Pasi Pertilä, Mikko Parviainen, Teemu Korhonen and Ari Visa Institute of Signal Proessing Tampere University of Tehnology, P.O.Box 553, FIN-330,

More information

Maximum Entropy and Exponential Families

Maximum Entropy and Exponential Families Maximum Entropy and Exponential Families April 9, 209 Abstrat The goal of this note is to derive the exponential form of probability distribution from more basi onsiderations, in partiular Entropy. It

More information

Complexity of Regularization RBF Networks

Complexity of Regularization RBF Networks Complexity of Regularization RBF Networks Mark A Kon Department of Mathematis and Statistis Boston University Boston, MA 02215 mkon@buedu Leszek Plaskota Institute of Applied Mathematis University of Warsaw

More information

Eigenvalues of tridiagonal matrix using Strum Sequence and Gerschgorin theorem

Eigenvalues of tridiagonal matrix using Strum Sequence and Gerschgorin theorem Eigenvalues o tridiagonal matrix using Strum Sequene and Gershgorin theorem T.D.Roopamala Department o Computer Siene and Engg., Sri Jayahamarajendra College o Engineering Mysore INDIA roopa_td@yahoo.o.in

More information

Microeconomic Theory I Assignment #7 - Answer key

Microeconomic Theory I Assignment #7 - Answer key Miroeonomi Theory I Assignment #7 - Answer key. [Menu priing in monopoly] Consider the example on seond-degree prie disrimination (see slides 9-93). To failitate your alulations, assume H = 5, L =, and

More information

Modeling Probabilistic Measurement Correlations for Problem Determination in Large-Scale Distributed Systems

Modeling Probabilistic Measurement Correlations for Problem Determination in Large-Scale Distributed Systems 009 9th IEEE International Conferene on Distributed Computing Systems Modeling Probabilisti Measurement Correlations for Problem Determination in Large-Sale Distributed Systems Jing Gao Guofei Jiang Haifeng

More information

Optimization of Statistical Decisions for Age Replacement Problems via a New Pivotal Quantity Averaging Approach

Optimization of Statistical Decisions for Age Replacement Problems via a New Pivotal Quantity Averaging Approach Amerian Journal of heoretial and Applied tatistis 6; 5(-): -8 Published online January 7, 6 (http://www.sienepublishinggroup.om/j/ajtas) doi:.648/j.ajtas.s.65.4 IN: 36-8999 (Print); IN: 36-96 (Online)

More information

State Diagrams. Margaret M. Fleck. 14 November 2011

State Diagrams. Margaret M. Fleck. 14 November 2011 State Diagrams Margaret M. Flek 14 November 2011 These notes over state diagrams. 1 Introdution State diagrams are a type of direted graph, in whih the graph nodes represent states and labels on the graph

More information

INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING Volume 2, No 4, 2012

INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING Volume 2, No 4, 2012 INTERNATIONAL JOURNAL OF CIVIL AND STRUCTURAL ENGINEERING Volume, No 4, 01 Copyright 010 All rights reserved Integrated Publishing servies Researh artile ISSN 0976 4399 Strutural Modelling of Stability

More information

Quantum secret sharing without entanglement

Quantum secret sharing without entanglement Quantum seret sharing without entanglement Guo-Ping Guo, Guang-Can Guo Key Laboratory of Quantum Information, University of Siene and Tehnology of China, Chinese Aademy of Sienes, Hefei, Anhui, P.R.China,

More information

arxiv: v2 [math.pr] 9 Dec 2016

arxiv: v2 [math.pr] 9 Dec 2016 Omnithermal Perfet Simulation for Multi-server Queues Stephen B. Connor 3th Deember 206 arxiv:60.0602v2 [math.pr] 9 De 206 Abstrat A number of perfet simulation algorithms for multi-server First Come First

More information

Chapter 3. Table of content Chapter 1: Switching Algebra Chapter 2: Logical Levels, Timing & Delays

Chapter 3. Table of content Chapter 1: Switching Algebra Chapter 2: Logical Levels, Timing & Delays hapter 3 Dr.-ng. Stefan Werner Tale of ontent hapter 1: Swithing lgera hapter 2: Logial Levels, Timing & Delays hapter 3: Karnaugh-Veith-Maps hapter 4: ominational iruit Design hapter 5: Lathes and Flip

More information

On the Bit Error Probability of Noisy Channel Networks With Intermediate Node Encoding I. INTRODUCTION

On the Bit Error Probability of Noisy Channel Networks With Intermediate Node Encoding I. INTRODUCTION 5188 IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 54, NO. 11, NOVEMBER 2008 [8] A. P. Dempster, N. M. Laird, and D. B. Rubin, Maximum likelihood estimation from inomplete data via the EM algorithm, J.

More information

CMSC 451: Lecture 9 Greedy Approximation: Set Cover Thursday, Sep 28, 2017

CMSC 451: Lecture 9 Greedy Approximation: Set Cover Thursday, Sep 28, 2017 CMSC 451: Leture 9 Greedy Approximation: Set Cover Thursday, Sep 28, 2017 Reading: Chapt 11 of KT and Set 54 of DPV Set Cover: An important lass of optimization problems involves overing a ertain domain,

More information

Two-Level Minimization

Two-Level Minimization Two-Level Minimization Logi Ciruits Design Seminars WS2010/2011, Leture 5 Ing. Petr Fišer, Ph.D. Department of Digital Design Faulty of Information Tehnology Czeh Tehnial University in Prague Evropský

More information

Sensitivity Analysis in Markov Networks

Sensitivity Analysis in Markov Networks Sensitivity Analysis in Markov Networks Hei Chan and Adnan Darwihe Computer Siene Department University of California, Los Angeles Los Angeles, CA 90095 {hei,darwihe}@s.ula.edu Abstrat This paper explores

More information

c-perfect Hashing Schemes for Binary Trees, with Applications to Parallel Memories

c-perfect Hashing Schemes for Binary Trees, with Applications to Parallel Memories -Perfet Hashing Shemes for Binary Trees, with Appliations to Parallel Memories (Extended Abstrat Gennaro Cordaso 1, Alberto Negro 1, Vittorio Sarano 1, and Arnold L.Rosenberg 2 1 Dipartimento di Informatia

More information

Normative and descriptive approaches to multiattribute decision making

Normative and descriptive approaches to multiattribute decision making De. 009, Volume 8, No. (Serial No.78) China-USA Business Review, ISSN 57-54, USA Normative and desriptive approahes to multiattribute deision making Milan Terek (Department of Statistis, University of

More information

The Laws of Acceleration

The Laws of Acceleration The Laws of Aeleration The Relationships between Time, Veloity, and Rate of Aeleration Copyright 2001 Joseph A. Rybzyk Abstrat Presented is a theory in fundamental theoretial physis that establishes the

More information

Tight bounds for selfish and greedy load balancing

Tight bounds for selfish and greedy load balancing Tight bounds for selfish and greedy load balaning Ioannis Caragiannis Mihele Flammini Christos Kaklamanis Panagiotis Kanellopoulos Lua Mosardelli Deember, 009 Abstrat We study the load balaning problem

More information

Supplementary Materials

Supplementary Materials Supplementary Materials Neural population partitioning and a onurrent brain-mahine interfae for sequential motor funtion Maryam M. Shanehi, Rollin C. Hu, Marissa Powers, Gregory W. Wornell, Emery N. Brown

More information

SINCE Zadeh s compositional rule of fuzzy inference

SINCE Zadeh s compositional rule of fuzzy inference IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 14, NO. 6, DECEMBER 2006 709 Error Estimation of Perturbations Under CRI Guosheng Cheng Yuxi Fu Abstrat The analysis of stability robustness of fuzzy reasoning

More information

Wavetech, LLC. Ultrafast Pulses and GVD. John O Hara Created: Dec. 6, 2013

Wavetech, LLC. Ultrafast Pulses and GVD. John O Hara Created: Dec. 6, 2013 Ultrafast Pulses and GVD John O Hara Created: De. 6, 3 Introdution This doument overs the basi onepts of group veloity dispersion (GVD) and ultrafast pulse propagation in an optial fiber. Neessarily, it

More information

Branch and Price for Multi-Agent Plan Recognition

Branch and Price for Multi-Agent Plan Recognition Proeedings of the Twenty-Fifth AAAI Conferene on Artifiial Intelligene Branh and Prie for Multi-Agent Plan Reognition Bikramjit Banerjee and Landon Kraemer Shool of Computing University of Southern Mississippi

More information

Product Policy in Markets with Word-of-Mouth Communication. Technical Appendix

Product Policy in Markets with Word-of-Mouth Communication. Technical Appendix rodut oliy in Markets with Word-of-Mouth Communiation Tehnial Appendix August 05 Miro-Model for Inreasing Awareness In the paper, we make the assumption that awareness is inreasing in ustomer type. I.e.,

More information

A Heuristic Approach for Design and Calculation of Pressure Distribution over Naca 4 Digit Airfoil

A Heuristic Approach for Design and Calculation of Pressure Distribution over Naca 4 Digit Airfoil IOSR Journal of Engineering (IOSRJEN) ISSN (e): 2250-3021, ISSN (p): 2278-8719 PP 11-15 www.iosrjen.org A Heuristi Approah for Design and Calulation of Pressure Distribution over Naa 4 Digit Airfoil G.

More information

Computer Science 786S - Statistical Methods in Natural Language Processing and Data Analysis Page 1

Computer Science 786S - Statistical Methods in Natural Language Processing and Data Analysis Page 1 Computer Siene 786S - Statistial Methods in Natural Language Proessing and Data Analysis Page 1 Hypothesis Testing A statistial hypothesis is a statement about the nature of the distribution of a random

More information

Statistical physics analysis of the computational complexity of solving random satisfiability problems using backtrack algorithms

Statistical physics analysis of the computational complexity of solving random satisfiability problems using backtrack algorithms Eur. Phys. J. B, 55 53 ) THE EUROPEAN PHYSICAL JOURNAL B EDP Sienes Soietà Italiana di Fisia Springer-Verlag Statistial physis analysis of the omputational omplexity of solving random satisfiability problems

More information

Evaluation of effect of blade internal modes on sensitivity of Advanced LIGO

Evaluation of effect of blade internal modes on sensitivity of Advanced LIGO Evaluation of effet of blade internal modes on sensitivity of Advaned LIGO T0074-00-R Norna A Robertson 5 th Otober 00. Introdution The urrent model used to estimate the isolation ahieved by the quadruple

More information

A NETWORK SIMPLEX ALGORITHM FOR THE MINIMUM COST-BENEFIT NETWORK FLOW PROBLEM

A NETWORK SIMPLEX ALGORITHM FOR THE MINIMUM COST-BENEFIT NETWORK FLOW PROBLEM NETWORK SIMPLEX LGORITHM FOR THE MINIMUM COST-BENEFIT NETWORK FLOW PROBLEM Cen Çalışan, Utah Valley University, 800 W. University Parway, Orem, UT 84058, 801-863-6487, en.alisan@uvu.edu BSTRCT The minimum

More information

Average Rate Speed Scaling

Average Rate Speed Scaling Average Rate Speed Saling Nikhil Bansal David P. Bunde Ho-Leung Chan Kirk Pruhs May 2, 2008 Abstrat Speed saling is a power management tehnique that involves dynamially hanging the speed of a proessor.

More information

Buckling loads of columns of regular polygon cross-section with constant volume and clamped ends

Buckling loads of columns of regular polygon cross-section with constant volume and clamped ends 76 Bukling loads of olumns of regular polygon ross-setion with onstant volume and lamped ends Byoung Koo Lee Dept. of Civil Engineering, Wonkwang University, Iksan, Junuk, 7-79, Korea Email: kleest@wonkwang.a.kr

More information

UTC. Engineering 329. Proportional Controller Design. Speed System. John Beverly. Green Team. John Beverly Keith Skiles John Barker.

UTC. Engineering 329. Proportional Controller Design. Speed System. John Beverly. Green Team. John Beverly Keith Skiles John Barker. UTC Engineering 329 Proportional Controller Design for Speed System By John Beverly Green Team John Beverly Keith Skiles John Barker 24 Mar 2006 Introdution This experiment is intended test the variable

More information

Word of Mass: The Relationship between Mass Media and Word-of-Mouth

Word of Mass: The Relationship between Mass Media and Word-of-Mouth Word of Mass: The Relationship between Mass Media and Word-of-Mouth Roman Chuhay Preliminary version Marh 6, 015 Abstrat This paper studies the optimal priing and advertising strategies of a firm in the

More information

Announcements. Office Hours Swap: OH schedule has been updated to reflect this.

Announcements. Office Hours Swap: OH schedule has been updated to reflect this. SA Solving Announements Offie Hours Swap: Zavain has offie hours from 4-6PM toay in builing 460, room 040A. Rose has offie hours tonight from 7-9PM in Gates B26B. Keith has offie hours hursay from 2-4PM

More information

Methods of evaluating tests

Methods of evaluating tests Methods of evaluating tests Let X,, 1 Xn be i.i.d. Bernoulli( p ). Then 5 j= 1 j ( 5, ) T = X Binomial p. We test 1 H : p vs. 1 1 H : p>. We saw that a LRT is 1 if t k* φ ( x ) =. otherwise (t is the observed

More information

Counting Idempotent Relations

Counting Idempotent Relations Counting Idempotent Relations Beriht-Nr. 2008-15 Florian Kammüller ISSN 1436-9915 2 Abstrat This artile introdues and motivates idempotent relations. It summarizes haraterizations of idempotents and their

More information

Danielle Maddix AA238 Final Project December 9, 2016

Danielle Maddix AA238 Final Project December 9, 2016 Struture and Parameter Learning in Bayesian Networks with Appliations to Prediting Breast Caner Tumor Malignany in a Lower Dimension Feature Spae Danielle Maddix AA238 Final Projet Deember 9, 2016 Abstrat

More information

Simplified Buckling Analysis of Skeletal Structures

Simplified Buckling Analysis of Skeletal Structures Simplified Bukling Analysis of Skeletal Strutures B.A. Izzuddin 1 ABSRAC A simplified approah is proposed for bukling analysis of skeletal strutures, whih employs a rotational spring analogy for the formulation

More information

MultiPhysics Analysis of Trapped Field in Multi-Layer YBCO Plates

MultiPhysics Analysis of Trapped Field in Multi-Layer YBCO Plates Exerpt from the Proeedings of the COMSOL Conferene 9 Boston MultiPhysis Analysis of Trapped Field in Multi-Layer YBCO Plates Philippe. Masson Advaned Magnet Lab *7 Main Street, Bldg. #4, Palm Bay, Fl-95,

More information

A Functional Representation of Fuzzy Preferences

A Functional Representation of Fuzzy Preferences Theoretial Eonomis Letters, 017, 7, 13- http://wwwsirporg/journal/tel ISSN Online: 16-086 ISSN Print: 16-078 A Funtional Representation of Fuzzy Preferenes Susheng Wang Department of Eonomis, Hong Kong

More information

ONLINE APPENDICES for Cost-Effective Quality Assurance in Crowd Labeling

ONLINE APPENDICES for Cost-Effective Quality Assurance in Crowd Labeling ONLINE APPENDICES for Cost-Effetive Quality Assurane in Crowd Labeling Jing Wang Shool of Business and Management Hong Kong University of Siene and Tehnology Clear Water Bay Kowloon Hong Kong jwang@usthk

More information

23.1 Tuning controllers, in the large view Quoting from Section 16.7:

23.1 Tuning controllers, in the large view Quoting from Section 16.7: Lesson 23. Tuning a real ontroller - modeling, proess identifiation, fine tuning 23.0 Context We have learned to view proesses as dynami systems, taking are to identify their input, intermediate, and output

More information

Name: ID: *************** write your answer in the right column *************** Total /30

Name: ID: *************** write your answer in the right column *************** Total /30 Course name: Network I Exam numer: Midterm II Model Answer Course Code: CNE 304 Exam Date: 21/11/2011 Leturer: Dr. Ahmed ElShafee Time Allowed: 90 minutes Name: ID: *************** write your answer in

More information

Hankel Optimal Model Order Reduction 1

Hankel Optimal Model Order Reduction 1 Massahusetts Institute of Tehnology Department of Eletrial Engineering and Computer Siene 6.245: MULTIVARIABLE CONTROL SYSTEMS by A. Megretski Hankel Optimal Model Order Redution 1 This leture overs both

More information

A Queueing Model for Call Blending in Call Centers

A Queueing Model for Call Blending in Call Centers A Queueing Model for Call Blending in Call Centers Sandjai Bhulai and Ger Koole Vrije Universiteit Amsterdam Faulty of Sienes De Boelelaan 1081a 1081 HV Amsterdam The Netherlands E-mail: {sbhulai, koole}@s.vu.nl

More information

Aircraft CAS Design with Input Saturation Using Dynamic Model Inversion

Aircraft CAS Design with Input Saturation Using Dynamic Model Inversion International Journal of Control, Automation, and Systems Vol., No. 3, September 003 35 Airraft CAS Design with Input Saturation sing Dynami Model Inversion Sangsoo Lim and Byoung Soo Kim Abstrat: This

More information

SAT-based Combinational Equivalence Checking

SAT-based Combinational Equivalence Checking SAT-based Combinational Equivalence Checking Zhuo Huang zhuang@cise.ufl.edu Prabhat Mishra prabhat@cise.ufl.edu CISE Technical Report #05-007 Department of Computer and Information Science and Engineering,

More information

Edexcel GCSE Maths Foundation Skills Book Ratio, proportion and rates of change 1

Edexcel GCSE Maths Foundation Skills Book Ratio, proportion and rates of change 1 Guidane on the use of odes for this mark sheme ethod mark A C P ao oe ft Auray mark ark awarded independent of method Communiation mark Proof, proess or justifiation mark Corret answer only Or equivalent

More information

Relative Maxima and Minima sections 4.3

Relative Maxima and Minima sections 4.3 Relative Maxima and Minima setions 4.3 Definition. By a ritial point of a funtion f we mean a point x 0 in the domain at whih either the derivative is zero or it does not exists. So, geometrially, one

More information

Nonseparable Space-Time Covariance Models: Some Parametric Families 1

Nonseparable Space-Time Covariance Models: Some Parametric Families 1 Mathematial Geology, ol. 34, No. 1, January 2002 ( C 2002) Nonseparale Spae-Time Covariane Models: Some Parametri Families 1 S. De Iao, 2 D. E. Myers, 3 and D. Posa 4,5 By extending the produt and produt

More information

Compression Members Local Buckling and Section Classification

Compression Members Local Buckling and Section Classification Compression Memers Loal Bukling and Setion Classifiation Summary: Strutural setions may e onsidered as an assemly of individual plate elements. Plate elements may e internal (e.g. the wes of open eams

More information

max min z i i=1 x j k s.t. j=1 x j j:i T j

max min z i i=1 x j k s.t. j=1 x j j:i T j AM 221: Advaned Optimization Spring 2016 Prof. Yaron Singer Leture 22 April 18th 1 Overview In this leture, we will study the pipage rounding tehnique whih is a deterministi rounding proedure that an be

More information

A simple expression for radial distribution functions of pure fluids and mixtures

A simple expression for radial distribution functions of pure fluids and mixtures A simple expression for radial distribution funtions of pure fluids and mixtures Enrio Matteoli a) Istituto di Chimia Quantistia ed Energetia Moleolare, CNR, Via Risorgimento, 35, 56126 Pisa, Italy G.

More information

An Integrated Architecture of Adaptive Neural Network Control for Dynamic Systems

An Integrated Architecture of Adaptive Neural Network Control for Dynamic Systems An Integrated Arhiteture of Adaptive Neural Network Control for Dynami Systems Robert L. Tokar 2 Brian D.MVey2 'Center for Nonlinear Studies, 2Applied Theoretial Physis Division Los Alamos National Laboratory,

More information

The simulation analysis of the bridge rectifier continuous operation in AC circuit

The simulation analysis of the bridge rectifier continuous operation in AC circuit Computer Appliations in Eletrial Engineering Vol. 4 6 DOI 8/j.8-448.6. The simulation analysis of the bridge retifier ontinuous operation in AC iruit Mirosław Wiślik, Paweł Strząbała Kiele University of

More information

Bypassing Space Explosion in High-Speed Regular Expression Matching

Bypassing Space Explosion in High-Speed Regular Expression Matching Bypassing Spae Explosion in High-Speed Regular Expression Mathing Jignesh Patel Alex X. Liu Eri Torng Astrat Network intrusion detetion and prevention systems ommonly use regular expression (RE) signatures

More information

General Closed-form Analytical Expressions of Air-gap Inductances for Surfacemounted Permanent Magnet and Induction Machines

General Closed-form Analytical Expressions of Air-gap Inductances for Surfacemounted Permanent Magnet and Induction Machines General Closed-form Analytial Expressions of Air-gap Indutanes for Surfaemounted Permanent Magnet and Indution Mahines Ronghai Qu, Member, IEEE Eletroni & Photoni Systems Tehnologies General Eletri Company

More information

Canimals. borrowed, with thanks, from Malaspina University College/Kwantlen University College

Canimals. borrowed, with thanks, from Malaspina University College/Kwantlen University College Canimals borrowed, with thanks, from Malaspina University College/Kwantlen University College http://ommons.wikimedia.org/wiki/file:ursus_maritimus_steve_amstrup.jpg Purpose Investigate the rate of heat

More information

Adaptive neuro-fuzzy inference system-based controllers for smart material actuator modelling

Adaptive neuro-fuzzy inference system-based controllers for smart material actuator modelling Adaptive neuro-fuzzy inferene system-based ontrollers for smart material atuator modelling T L Grigorie and R M Botez Éole de Tehnologie Supérieure, Montréal, Quebe, Canada The manusript was reeived on

More information

A model for measurement of the states in a coupled-dot qubit

A model for measurement of the states in a coupled-dot qubit A model for measurement of the states in a oupled-dot qubit H B Sun and H M Wiseman Centre for Quantum Computer Tehnology Centre for Quantum Dynamis Griffith University Brisbane 4 QLD Australia E-mail:

More information

LOGISTIC REGRESSION IN DEPRESSION CLASSIFICATION

LOGISTIC REGRESSION IN DEPRESSION CLASSIFICATION LOGISIC REGRESSIO I DEPRESSIO CLASSIFICAIO J. Kual,. V. ran, M. Bareš KSE, FJFI, CVU v Praze PCP, CS, 3LF UK v Praze Abstrat Well nown logisti regression and the other binary response models an be used

More information

Chapter 3 Church-Turing Thesis. CS 341: Foundations of CS II

Chapter 3 Church-Turing Thesis. CS 341: Foundations of CS II CS 341: Foundations of CS II Marvin K. Nakayama Computer Siene Department New Jersey Institute of Tehnology Newark, NJ 07102 CS 341: Chapter 3 3-2 Contents Turing Mahines Turing-reognizable Turing-deidable

More information

Likelihood-confidence intervals for quantiles in Extreme Value Distributions

Likelihood-confidence intervals for quantiles in Extreme Value Distributions Likelihood-onfidene intervals for quantiles in Extreme Value Distributions A. Bolívar, E. Díaz-Franés, J. Ortega, and E. Vilhis. Centro de Investigaión en Matemátias; A.P. 42, Guanajuato, Gto. 36; Méxio

More information

1. INTRODUCTION. l t t r. h t h w. t f t w. h p h s. d b D F. b b d c. L D s

1. INTRODUCTION. l t t r. h t h w. t f t w. h p h s. d b D F. b b d c. L D s Rapid Assessment of Seismi Safety of Elevated ater Tanks with FRAME Staging 1. NTRODUCTON 1.1 ntrodution ater tanks are lifeline items in the aftermath of earthquakes. The urrent pratie of designing elevated

More information

A Characterization of Wavelet Convergence in Sobolev Spaces

A Characterization of Wavelet Convergence in Sobolev Spaces A Charaterization of Wavelet Convergene in Sobolev Spaes Mark A. Kon 1 oston University Louise Arakelian Raphael Howard University Dediated to Prof. Robert Carroll on the oasion of his 70th birthday. Abstrat

More information

Calibration of Piping Assessment Models in the Netherlands

Calibration of Piping Assessment Models in the Netherlands ISGSR 2011 - Vogt, Shuppener, Straub & Bräu (eds) - 2011 Bundesanstalt für Wasserbau ISBN 978-3-939230-01-4 Calibration of Piping Assessment Models in the Netherlands J. Lopez de la Cruz & E.O.F. Calle

More information

Sensor management for PRF selection in the track-before-detect context

Sensor management for PRF selection in the track-before-detect context Sensor management for PRF seletion in the tra-before-detet ontext Fotios Katsilieris, Yvo Boers, and Hans Driessen Thales Nederland B.V. Haasbergerstraat 49, 7554 PA Hengelo, the Netherlands Email: {Fotios.Katsilieris,

More information

Appendix A Market-Power Model of Business Groups. Robert C. Feenstra Deng-Shing Huang Gary G. Hamilton Revised, November 2001

Appendix A Market-Power Model of Business Groups. Robert C. Feenstra Deng-Shing Huang Gary G. Hamilton Revised, November 2001 Appendix A Market-Power Model of Business Groups Roert C. Feenstra Deng-Shing Huang Gary G. Hamilton Revised, Novemer 200 Journal of Eonomi Behavior and Organization, 5, 2003, 459-485. To solve for the

More information

QCLAS Sensor for Purity Monitoring in Medical Gas Supply Lines

QCLAS Sensor for Purity Monitoring in Medical Gas Supply Lines DOI.56/sensoren6/P3. QLAS Sensor for Purity Monitoring in Medial Gas Supply Lines Henrik Zimmermann, Mathias Wiese, Alessandro Ragnoni neoplas ontrol GmbH, Walther-Rathenau-Str. 49a, 7489 Greifswald, Germany

More information

11.1 Polynomial Least-Squares Curve Fit

11.1 Polynomial Least-Squares Curve Fit 11.1 Polynomial Least-Squares Curve Fit A. Purpose This subroutine determines a univariate polynomial that fits a given disrete set of data in the sense of minimizing the weighted sum of squares of residuals.

More information

Design and Development of Three Stages Mixed Sampling Plans for Variable Attribute Variable Quality Characteristics

Design and Development of Three Stages Mixed Sampling Plans for Variable Attribute Variable Quality Characteristics International Journal of Statistis and Systems ISSN 0973-2675 Volume 12, Number 4 (2017), pp. 763-772 Researh India Publiations http://www.ripubliation.om Design and Development of Three Stages Mixed Sampling

More information

Nozzle Fuzzy Controller of Agricultural Spraying Robot Aiming Toward Crop Rows

Nozzle Fuzzy Controller of Agricultural Spraying Robot Aiming Toward Crop Rows Nozzle Fuzzy ontroller of Agriultural Spraying Root Aiming Toward rop Rows Jianqiang Ren Department of omputer, Langfang Teahers ollege, Langfang, Heei Provine, P.R. hina 065000, Tel.:+86-316-681780 renjianqiang@163.om

More information

EXACT TRAVELLING WAVE SOLUTIONS FOR THE GENERALIZED KURAMOTO-SIVASHINSKY EQUATION

EXACT TRAVELLING WAVE SOLUTIONS FOR THE GENERALIZED KURAMOTO-SIVASHINSKY EQUATION Journal of Mathematial Sienes: Advanes and Appliations Volume 3, 05, Pages -3 EXACT TRAVELLING WAVE SOLUTIONS FOR THE GENERALIZED KURAMOTO-SIVASHINSKY EQUATION JIAN YANG, XIAOJUAN LU and SHENGQIANG TANG

More information

On Industry Structure and Firm Conduct in Long Run Equilibrium

On Industry Structure and Firm Conduct in Long Run Equilibrium www.siedu.a/jms Journal of Management and Strategy Vol., No. ; Deember On Industry Struture and Firm Condut in Long Run Equilibrium Prof. Jean-Paul Chavas Department of Agriultural and Applied Eonomis

More information