DD representation, the equivalence chec or specication and implementation translates to the chec whether the corresponding DDs are identical. The most
|
|
- Erin Reynolds
- 6 years ago
- Views:
Transcription
1 Technical Report 02, Albert-Ludwigs-University, Freiburg, May 998. Word-Level Decision Diagrams, WLCDs and Division Christoph Scholl Bernd Becer Thomas M. Weis Institute o Computer Science Albert{Ludwigs{University D 790 Freiburg im Breisgau, Germany <name>@inormati.uni-reiburg.de Abstract Several types o Decision Diagrams (DDs) have been proposed or the verication o Integrated Circuits. Recently, word-level DDs lie bmds, *bmds, hdds, *bmds and *phdds have been attracting more and more interest, e.g., by using *bmds and *phdds it was or the rst time possible to ormally veriy integer multipliers and oating point multipliers o \signicant" bitlengths, respectively. On the other hand, it has been unnown, whether division, the operation inverse to multiplication, can be eciently represented by some type o word-level DDs. In this paper we show that the representational power o any word-level DD is too wea to eciently represent integer division. Thus, neither a clever choice o the variable ordering, the decomposition type or the edge weights, can lead to a polynomial DD size or division. For the proo we introduce Word-Level Linear Combination Diagrams (wlcds), a DD, which may be viewed as a \generic" word-level DD. We derive an exponential lower bound on the wlcd representation size or integer dividers and show how this bound transers to all other word-level DDs. Introduction One o the most important tass during the design o Integrated Circuits is the verication o an implemented circuit, i.e., the chec whether the implementation ullls its specication. In the last ew years several methods based on Decision Diagrams (DDs) have been proposed [5, 2, 5] to perorm verication. The idea is to transorm both, implementation and specication o a combinational circuit, into a DD. Then, due to the canonicity o the
2 DD representation, the equivalence chec or specication and implementation translates to the chec whether the corresponding DDs are identical. The most popular data structure in this context are Ordered Binary Decision Diagrams (obdds) [3]. They were applied successully e.g. to the verication o control logic and integer adders. But there are unctions o high practical relevance, which cannot be represented eciently by obdds. Already in [3] and [4] Bryant proved that obdd representations or integer multipliers are o exponential size. Several other types o DDs were dened to overcome the limitations o obdds, such as Ordered Functional Decision Diagrams (odds) [3], Ordered Kronecer Functional Decision Diagrams (odds) [], Multi{Terminal Binary Decision Diagrams (mtbdds) [9] (also called Algebraic Decision Diagrams (adds) [] and Edge{valued Binary Decision Diagrams (evbdds) [4]. But the rst DDs to represent integer multiplication eciently were Binary Moment Diagrams (bmds) and Multiplicative bmds (*bmds) introduced in [6]. Lie mtbdds and evbdds, also bmds and *bmds are word-level DDs, i.e. they represent integer-valued unctions : 0; g n! Z. To urther improve on the representational power o bmds, several other word-level DD types have been introduced, e.g. Hybrid Decision Diagrams (hdds) [8] and Kronecer *bmds (*bmds) [0]. Recently Chen and Bryant dened a new data structure called Multiplicative Power Hybrid Decision Diagrams (*phdds) [7], which is able to represent not only integer multiplication but also oating point multiplication eciently. Until now it was not nown, whether the word-level DDs mentioned above are also able to represent division eciently. Recently Naanishi [6] made a rst step by showing that *bmds cannot represent integer division eciently. The proo is technically complicated, it is based on ooling set arguments similar to the orginal proo or multiplication by Bryant and has to tae into account the edge values in the *bmd representation. Consequently, as already mentioned, in this orm it only wors or *bmds. In this paper we prove that integer division cannot be represented in polynomial size by any o the ordered word-level DDs mentioned in the literature until now. Even more interestingly, we prove that the concept o word-level DDs in general is too wea to result in polynomial size representations o division. For the proo we introduce a new data structure, the Word-Level Linear Combination Diagrams (wlcds). wlcds are a generalization o Waac's Parity Ordered Binary Decision Diagrams (pobdds) [7] to the word level. It turns out that wlcds can be viewed as a \generic" ordered word-level DD in the ollowing sense: Each ordered word-level DD can be \embedded into" wlcds such that a DD with nodes is transormed into a wlcd representing the same unction with the same number o nodes. Thus, a lower bound on the size o a wlcd is also a lower bound on the size o all other ordered word-level diagrams. We apply this idea to integer division by deriving an exponential lower bound on the size o wlcds representing integer divison (regardless o the chosen variable order). For wlcds 2
3 lower bounds can be obtained by consideration o the ran o a communication matrix which is constructed rom the unction tables o several coactors. It ollows that bothering details concerning e.g. edge values have not to be taen into account to derive the lower bound in our proo. On the other hand, due to the properties o wlcds we obtain an exponential lower bound result, valid or all ordered word-level DD types. The paper is structured as ollows. In Section 2 we provide basics on word-level DDs which will be necessary or the understanding o the paper. wlcds and their relationship to existing word-level DDs are introduced in Section 3. Furthermore, an algebraic characterization o the wlcd complexity is given which leads to the ran considerations o certain coactor matrices. In Section 4 the lower bound or division is derived. We nish with conclusions and perspectives o urther wor in Section 5. 2 Preliminaries: Word-Level Decision Diagrams In this section we give a short review o ordered word-level DDs, data structures used or the representation o so-called Pseudo Boolean unctions, i.e. unctions rom a Boolean domain to the integers or rational numbers. In general, DDs are graph{based representations, where at each (non{terminal) node (labeled with a variable x) a decomposition o the unction (represented by this node) into two subunctions (the low{unction and the high{ unction) is perormed: Denition A word-level DD is a rooted directed acyclic graph G = (V; E) with non empty vertex set V containing two types o vertices, non-terminal and terminal vertices. A non-terminal vertex v has as label a variable index(v) 2 x ; : : : ; x n g and two children low(v); high(v) 2 V. A terminal vertex v is labeled with a value value(v) 2 Z. For the purpose o this paper, we are only interested in ordered DDs, i.e. DDs, where the variables occur in the same order on all paths o the DD. More precisely, this means: Denition 2 A DD is ordered i there is a xed order : ; : : : ; ng! x ; : : : ; x n g such that or any non-terminal vertex v the ollowing holds: index(low(v)) = () with >? (index(v)) (index(high(v)) = (q) with q >? (index(v))) as long as low(v) (high(v)) is also a non-terminal vertex. Based on these general denitions we now consider dierent decomposition types and shortly discuss resulting word-level DDs and corresponding evaluation rules. (For a survey on word-level DDs and more details see also [2].) 3
4 2. Decomposition types and evaluation rules In word-level diagrams the unction v : 0; g n! Q represented by a non{terminal node v, which is labeled by variable x i, is decomposed into two subunctions, both independent o variable x i. Depending on the decomposition type these subunctions are combined rom the coactors ( v ) xi = v (x ; : : : ; x i? ; 0; x i+ ; : : : ; x n ) and ( v ) xi = v (x ; : : : ; x i? ; ; x i+ ; : : : ; x n ) in dierent ways. DDs as dened in literature dier in the way they use decomposition types. Decomposition types can be dened by the set Z 2;2 o non{singular 2 2 matrices over Z [8]. The most important decomposition types are Shannon decomposition, positive Davio decomposition and negative Davio decomposition. The Shannon decomposition is used in mtbdds [9] and evbdds [4], the positive Davio decomposition is used in bmds and *bmds [6]. In *bmds [0] and *phdds [7] Shannon decomposition, positive Davio and negative Davio decomposition are used. In hdds [8] six dierent decomposition types (including Shannon, positive and negative Davio decomposition) are used. Following [8] the matrices corresponding to Shannon, positive Davio and negative Davio decomposition, respectively, are and : 0?? These matrices dene how the unctions low(v) and high(v) represented by low(v) and high(v) are computed rom ( v ) xi and ( v ) xi. For the positive Davio decomposition, e.g., we have low(v) 0 (v ) = xi ;? ( v ) xi high(v) i.e., low(v) = ( v ) xi and high(v) = ( v ) xi? ( v ) xi. A terminal node v with value(v) = z represents the constant unction with unction value z. To evaluate the unction v represented by a non{terminal node v or x i = 0 or x i =, we have to reconstruct ( v ) xi or ( v ) xi rom low(v) and high(v). To do so, we mae use o the act, that the decomposition type matrices are non{singular: Since a decomposition type matrix A is non{singular, the inverse matrix A? exists and (v ) xi = A? low(v) ( v ) xi high(v) : () The inverse decomposition type matrices or Shannon, positive Davio and negative Davio decomposition, respectively, are 0 0 and : 0 0 4
5 For positive Davio decomposition, e.g., this means that ( v ) xi = low(v) and ( v ) xi = low(v) + high(v). 2.2 Additive and multiplicative edge values, negation edges Edge values are introduced to increase the amount o subgraph sharing when using integer{ valued terminal nodes. It has to be dierentiated between additive and multiplicative edge values. An edge with additive weight a and multiplicative weight m leading to node v represents the unction < (a; m); v >:= a + m v : (2) mtbdds, bmds and hdds use no edge values, evbdds use only additive weights, i.e., the multiplicative weight m is, *bmds use only multiplicative weights, i.e. a = 0. *bmds use both additive and multiplicative weights. *phdds use only multiplicative weights o orm (?) ne 2 w with ne 2 0; g and w 2 Z. (For reasons o memory eciency (?) ne 2 w is stored as an integer w and a bit ne representing a `negation edge' when ne =.) Now consider any ordered word-level DD with edge values. Then or each non{terminal node v there is a 0{edge labeled with edge weights (a low ; m low ) leading to node low(v) and a {edge labeled with edge weights (a high ; m high ) leading to node high(v). I in node v the decomposition type A =? a a 2 a 2 a 22 with inverse matrix A? =? a 0 a0 2 a 0 2 a0 22 is used, then using Equations and 2 the evaluation rule or this node is the ollowing: v = (? x i ) ( v ) xi + x i ( v ) xi = (? x i ) (a 0 (a low + m low low(v) ) + a 0 2(a high + m high high(v) )) + x i (a 0 2(a low + m low low(v) ) + a 0 22(a high + m high high(v) )) = (? x i ) ((a 0 a low + a 0 2a high ) + (a 0 m low low(v) ) + (a 0 2m high high(v) )) + x i ((a 0 2a low + a 0 22a high ) + (a 0 2m low low(v) ) + (a 0 22m high high(v) )): (3) In Section 3 we will use the \most general evaluation rule" o Equation 3 to analyze the relationship between the existing ordered word-level DDs and our new data structure called Word-Level Linear Combination Diagrams (wlcds). 3 Word-Level Linear Combination Diagrams In this section we dene Word-Level Linear Combination Diagrams (wlcds). wlcds are a generalization o pobdds dened by Waac [7] to the word-level. Whereas pobdds can represent only Boolean unctions, wlcds represent unctions : 0; g n! Q. wlcds are given by the ollowing denition: 5
6 u... 0 v x i 0 w(v,u ) w(v,u ) w(v,w ) w(v,w l ) u w w l Figure : Non-terminal vertex v o a wlcd. v is labeled by variable x i. The 0{edges o v are given by edges to nodes u ; : : : ; u and the {edges are given by edges to w ; : : : ; w l. Denition 3 A Word-Level Linear Combination Diagram (wlcd) is a rooted directed acyclic graph G = (V; E). I the wlcd is not empty, it contains exactly one sin labeled with and with no outgoing edges. The remaining nodes are called non{terminal nodes. A non-terminal vertex v is labeled by a variable index(v) 2 x ; : : : ; x n g. The outgoing edges o a non{terminal node v are partitioned into two sets: 0{edges(v) and {edges(v). At least one o these sets is not empty. All edges e are labeled by an edge weight w(e) 2 Q. A wlcd is ordered, i.e., as with DDs the variables occur in the same order on all paths o wlcd. The size o a wlcd is its number o nodes. The denition o a wlcd is illustrated by Figure. An empty wlcd represents the constant 0{unction, the sin o a non{empty wlcd represents the constant {unction. The unction v represented by a non{terminal node v labeled by variable x i with 0{edges(v) = (v; u ); : : : ; (v; u )g and {edges(v) = (v; w ); : : : ; (v; w l )g is dened by the ollowing evaluation rule: v := (?x i )(w(v; u ) u +: : :+w(v; u ) u )+x i (w(v; w ) w +: : :+w(v; w l ) wl ): (4) Similar to pobdds, also or wlcds ecient synthesis operations and an equivalence chec can be derived. We omit any urther details, rather we concentrate on the property o wlcds which is most important in this paper: Ordered word-level DDs can be `embedded into wlcds', i.e., i there is some word-level DD with nodes, we can easily construct a wlcd with the same number o nodes. This act is used to conclude lower bounds on the size o arbitrary word-level diagrams rom lower bounds on the size o wlcds. The computation o lower bounds on the size o wlcds can be done in an elegant way using arguments rom linear algebra. Beore coming to lower bounds we show how to embed word-level DDs in wlcds. 6
7 3. Relationship between wlcds and existing word-level DDs Here we prove that all ordered word-level DDs mentioned in the previous sections can be `embedded into wlcds'. To do so we proceed as ollows: A given word-level DD is transormed step by step into a wlcd. I the given DD contains terminals v with values value(v) dierent rom 0 and, these terminals are replaced by a terminal and the edge weights o all incoming edges o v are multiplied by value(v). I now there is more than one terminal with value, these terminals are replaced by a unique terminal with value. Edges to terminal 0 with additive weight a 6= 0 are replaced by edges to terminal with additive weight a and multiplicative weight 0. The 0{terminal is removed. All these steps do not change the unction represented by the diagram. Now in a bottom{up procedure or each non{terminal node v labeled with variable x i = index(v) representing a unction v the outgoing edges are replaced resulting in a wlcd{ node representing the same unction v. Suppose that the decomposition type used or node v is given by A =? a a 2 a 2 a 22 (with inverse matrix A? =? a 0 a0 2 a 0 2 a0 22 ) and the 0{edge is labeled with edge weights (a low ; m low ), the {edge is labeled with edge weights (a high ; m high ). Then the evaluation rule o Equation 3 gives a relation between low(v) and high(v) and v. A comparison with the evaluation rule or wlcds (see Equation 4) leads to the denition o the equivalent wlcd{node and its corresponding edges (let v one be the terminal with value ): 0{edges(v) = (v; v one ); (v; low(v)); (v; high(v))g, w(v; v one ) = a 0 a low + a 0 2a high, w(v; low(v)) = a 0 m low, w(v; high(v)) = a 0 2m high. {edges(v) = (v; v one ); (v; low(v)); (v; high(v))g, w(v; v one ) = a 0 2a low + a 0 22a high, w(v; low(v)) = a 0 2m low, w(v; high(v)) = a 0 22m high. The replacement is illustrated by Figure 2. Ater this bottom{up procedure, i there is a root edge with weight (a; m), the weights o the outgoing edges o the root are multiplied by m and an edge (root; v one ) with weight a is included into 0{edges(root) and {edges(root). Finally we obtain a wlcd representing the same unction as the original DD. We summarize: Theorem I the mtbdd, evbdd, bmd, *bmd, hdd, *bmd or *phdd or a unction : 0; g n! Z (or : 0; g n! Q or the case o *phdds) with variable order has nodes, then there also exists a wlcd with variable order representing with (at most) nodes. 7
8 ( a, ) low mlow x i 0 ( a, m high high ) a a a ' ' low 2a high x 0 i 0 0 m low a ' a ' m 2 low a ' 2m high a ' 22m high lo w h ig h a a a ' ' 2 low 22a high lo w h ig h Figure 2: Transormation into wlcd node x i 0 x i 0 lo w h ig h lo w h ig h Figure 3: Transormation o a positive Davio node into a wlcd node Example In Figure 3 the node replacement described to prove Theorem is illustrated or positive Davio decomposition without edge weights (i.e. the additive edge weights are 0 and multiplicative edge weights are ). For positive Davio decomposition the decomposition type matrix ist given by A =? a a 2 a 2 a 22 =? 0? rule can be simplied to, A? =? a 0 a0 2 a 0 2 a0 22 =? 0. Thus, the evaluation v = (? x i ) ((a 0 a low + a 0 2a high ) + (a 0 m low low(v) ) + (a 0 2m high high(v) )) + x i ((a 0 2a low + a 0 22a high ) + (a 0 2m low low(v) ) + (a 0 22m high high(v) )) = (? x i ) low(v) + x i ( low(v) + high(v) ): 3.2 An Algebraic Characterization o the wlcd Complexity In this subsection we give an algebraic characterization o the wlcd complexity, which we will use to prove lower bounds on the size o wlcds. We show, that the number o nodes in a wlcd cannot be smaller than the dimension o a certain vector space. 8
9 Consider the set o all unctions rom 0; g n to the rational numbers Map(0; g n ; Q) = : 0; g n! Qg. Dene addition on Map(0; g n ; Q) by (+g)(x ; : : : ; x n ) = (x ; : : : ; x n )+ g(x ; : : : ; x n ) and multiplication with a scalar w 2 Q by (w)(x ; : : : ; x n ) = w(x ; : : : ; x n ). It is easy to see, that Map(0; g n ; Q) together with addition and multiplication with scalars orms a vector space. Based on wlcds with xed variable order we will dene subspaces o the vector space Map(0; g n ; Q). W.l.o.g. we assume the natural variable order, i.e., : ; : : : ; ng! x ; : : : ; x n g with (i) = x i 8i 2 ; : : : ; ng. Given a wlcd B, consider or some 2 ; : : : ; ng the set o all wlcd{nodes, which are labeled with variable x or which are labeled with a variable x i with i > and which have an incoming edge rom a node labeled by a variable x j with j <. These nodes represent unctions o Map(0; g n ; Q). We denote this set o unctions by V B. O course, the vector space < V B B > which is generated by the unctions in V orms a subspace o Map(0; g n ; Q). Let be the unction represented by the wlcd B. We consider the ollowing set o coactors o : = j x =c ;:::;x? =c? jc ; : : : ; c? 2 0; gg: Again, < V V >, which is generated by the unctions in V, is a subspace o Map(0; gn ; Q). Now we investigate the relationship between the vector spaces < V B claim that < V >< V B > : > and < V >. We To prove this it is sucient to show, that each coactor j x =c ;:::;x? =c? 2 V is in < V B >. We consider all paths starting rom the root o B, which ulll the assignment x = c ; : : : ; x? = c?. Let v ; : : : ; v m be the nodes, which are reached by these paths and suppose that each node v r 8r 2 : : : mg is reached by i r dierent paths p (r) ; : : : ; p (r) i r. Let w (r) j be the product o all weights o edges on path p (r) j. Then according to the denition o wlcds and by induction on the ollowing holds: j x =c ;:::;x? =c? = i X j= w () j! v + : : : + Since 8 i m vi 2 V B, we conclude that j x =c ;:::;x? =c? o c ; : : : ; c? 2 0; g. Because o < V and since V B >< V B generates < V B > we have dim(< V > it holds Thus we obtain the ollowing lemma dim(< V B >) dim(< V >) B >) jv j: 9 i m X j= w (m) j! 2< V B vm : > or each choice
10 Lemma Let be any unction in Map(0; g n ; Q). Then dim(< V >) is a lower bound on the size o a wlcd or with respect to the natural variable ordering. In act, we can prove even a stronger result with similar arguments as in the proo o Waac [7] or pobdds: Theorem 2 A wlcd B with natural variable S order representing unction with a minimal n+ number o nodes, has exactly dim(< V = >) nodes. However, or the purposes o this paper we need only Lemma. 4 An Exponential Lower Bound or Division In this section we apply Lemma to derive an exponential lower bound on the size o wlcds (and thus o word-level DDs) representing integer divison. For our proo we use the ollowing notations and denitions concerning division: Two sets o variables, the a-variables A = a n? ; : : : ; a 0 g and the b-variables B = b n? ; : : : ; b 0 g, are considered. As usual, the binary representation o A and B is given by jjajj := 2 n? a n? + : : : a 0 and jjbjj := 2 n? b n? + : : : b 0 ; respectively. Then the integer division DIV is the Pseudo Boolean unction dened by DIV : 0; g n 0; g n! IN: jjajj (a n? ; : : : ; a 0 ; b n? ; : : : ; b 0 ) 7! jjbjj Beore we prove the exponential lower bound or wlcds with arbitrary variable orders we consider a restricted case which nicely demonstrates the idea o the proo and the proo technique. Then we turn to the proo o the general case which is slightly more complicated but wors along similar lines. 4. wlcds with Interleaved Variable Ordering For the restricted case we x the variable order in advance: it is given by the interleaved ordering (a n? ; b n? ; : : : ; a 0 ; b 0 ): 0
11 a a b b a 3 a 2 b 3 b * * 0 0 * 2 0 * * 4 2 * 5 2 * * * * * * * * * * Figure 4: Communication matrix o division or n = 4. Furthermore, we may assume that n is even. (For n odd, we embed an (n? ){bit divider into the n{bit divider by setting a n? = b n? = 0 and note that or an exponential lower bound (c n ) it holds (c n ) = (c n? ).) Following Lemma we now consider the set V n+ o coactors V n+ = j an? =c ;b n? =c 2 ;:::;a n 2 =c n? ;b n 2 =c n and show an exponential lower bound or dim(< V n+ >). j c ; : : : ; c n 2 0; gg To estimate dim(< V n+ >) we prove that a certain number o elements o V n+ is linearly independent. For that we consider a communication matrix whose rows are unction tables o the coactors o Vn+. The rows o the matrix are \numbered" by input combinations o the \upper hal" o the a- and b-variables. Analogously, the \lower hal" o the a- and b-variables denes the columns. For illustration see Figure 4, where we give the communication matrix or n = 4. (A star in the matrix means that the corresponding result o DIV is not dened (division by zero). Our proo is valid or all possible replacements o the stars.) The ran o this communication matrix is equal to dim(< V n+ >). Since we need only a lower bound on dim(< V n+ >), we may remove columns and rows in the matrix (thereby possibly reducing the ran o the resulting matrix). The idea now is to restrict to entries with constant values or the b-variables and to observe the result o the division or increasing values o a-variables. More precisely, we only eep rows where rom the b-variables exactly the least signicant upper b-variable b n is set, i.e. 2 b n? = 0; : : : ; b n 2 + = 0; b n 2 =
12 Analogously, only columns with b n = 0; : : : ; b 2? = 0; b 0 = are considered. Furthermore, the rows and columns with 0 or all a-inputs are removed. For our example with n = 4 the ollowing matrix remains: a 0 a 0 0 b b 0 a 3 a 2 b 3 b In general, we obtain a matrix M o size (2 n 2? ) (2 n 2? ). For the computation o the entries o M consider the set A HIGH := a n? ; : : : ; a n g and A LOW := a n ; : : : ; a 2 2? P 0 g, n? i.e. A HIGH (A LOW ) consists o the upper (lower) a-variables. Dene jja HIGH jj := i= n a i 2 i? n P 2 2 n 2 and jja LOW jj :=? i=0 a i 2 i. Now let m ij denote an entry o M. Then row i corresponds to an assignment jja HIGH jj and column j corresponds to an assignment jja LOW jj. The entry m ij is then given by jjahigh jj 2 n 2 + jja LOW jj m ij = : 2 n 2 + (Remember that (b n? ; : : : ; b n n ; : : : ; b 2 2? 0 ) = (0; : : : ; 0; ).) It ollows that the result o DIV cannot be determined by looing at the assignment or A HIGH, rather the \relation" between \corresponding" bits o A HIGH and A LOW is essential: For jja HIGH jj = jja LOW jj the result is obviously jja HIGH jj. For jja LOW jj < jja HIGH jj we have m ij = jja HIGH jj?. (m ij < jja HIGH jj? would imply jja HIGH jj 2 n jja LOW jj which cannot be true.) For jja LOW jj > jja HIGH jj we obtain m ij = jja HIGH jj. Thus, the resulting matrix has the ollowing orm: M = n 2? 2 2 n 2? 2 2 n 2? 2 2 n 2? To prove that this matrix M has ull ran we apply column transormations starting with a subtraction o the rst column rom all other columns. For the resulting matrix the submatrix consisting o the last 2 n 2? 2 columns and the last 2 n 2? 2 rows is an upper triangular matrix with 's on the diagonal. By additional column tranormations it ollows directly that the matrix has maximum ran 2 n 2?. Consequently, the ran o the original communication matrix and dim(< V n+ >) is in (2 n 2 ). We summarize: 2 C A :
13 Lemma 2 A wlcd with interleaved variable ordering representing unction DIV has at least size (2 n 2? ). 4.2 The General Case In the rest o this section we extend the above proo to wlcds with arbitrary variable order. Crucial or the proo in the case o the interleaved variable ordering was the act that the outcome o DIV was depending on the assignments or corresponding variables in the upper and lower hal o the a-variables. More precisely: For the interleaved variable ordering there are n=2 pairs o variables (a i ; a j ) with a i (a j ) is an upper (a lower) a-variable a i and a j are split between the rst and the second hal o the variable order and the dierence o the indices i and j is a constant oset n 2. For arbitrary variable orders this is not always the case or the oset n, but i one modies 2 the oset to a number n? p 2 0, one can always nd \enough" such pairs being separated by the considered variable ordering. A precise ormulation o this (purely combinatorial) property, its proo (and the application to lower bounds or multiplication) has already been given by Bryant in [4]. Using this property the proo or the interleaved variable ordering can be modied by specication o a \similar" communication submatrix. Consideration o the ran o this submatrix then leads to the ollowing result: Theorem 3 A wlcd or unction DIV has at least size 2 n 6? (regardless o the variable order). Proo: We now give details o the proo: At rst the notion \suitable" pairs (a i ; a j ) is precisely specied: To do so, we adopt the notions o [4] and give a short review on the main points as ar as they are necessary or our proo. Let and A U = a n? ; : : : ; a n 2 g A D = a n 2? ; : : : ; a 0 g be the sets o upper and lower a-variables, respectively. Given any variable order or the variables o A and B we dene two sets L and R. L contains the rst l variables in the variable order and R contains the remaining 2n? l variables. l is chosen such that ja \ Lj = ja \ Rj, i.e. in L and R we have the same number o a-variables. 3
14 We dene the sets o upper and lower a-variables contained in L and R: A UL := A U \ L; A DL := A D \ L; A UR := A U \ R; A DR := A D \ R: In sets Args p pairs o variables (a i ; a j ) 2 A U A D are grouped with constant distance i? j = n? p: For? n p 0 Args 2 2 p := (a n 2?p+i ; a i )j0 i < n + pg and or p n? 2 2 Args p := (a n 2 +i ; a p+i )j0 i < n? pg. 2 Moreover let Split p = Args p \ ((A UL A DR ) [ (A UR A DL )) be the subset o Args p containing the pairs split between L and R. Following [4] there is a p 0 with jsplit p0 j > n 8 : Split p0 contains our set o suitable pairs which we will use to obtain the lower bound or wlcds with general variable orders. For the remaining part o the proo we assume that? n p (The case p 0 n? can be handled in an analogous manner.) Furthermore, since jsplit 2 p 0 j > n we have 8 jsplit p0 \ (A UL A DR )j > n or jsplit 6 p 0 \ (A UR A DL )j > n. W.o.l.g. we assume that 6 jsplit p0 \ (A UL A DR )j > n 6 : As in the special case we consider a communication matrix. The rows are unction tables o the coactors with respect to the variables o L. To estimate the ran o this matrix we remove rows and columns and compute the ran o the remaining submatrix. Let be minimal with (a n 2?p 0+; a ) 2 Split p0. We eep only rows and columns with b =, b n 2?p 0+ = and b i = 0 or i 2 0; : : : ; n? g n n? p ; g. In addition all a i which do not occur as components in Split p0 are set to 0. Dene and jja HIGH jj := jja LOW jj = X (a i ;a i? n 2 +p 0 )2Split p0 \(A UL A DR ) X (a n 2?p 0 +i ;a i)2split p0 \(A UL A DR ) a i 2 i? n 2 +p 0? a i 2 i? By varying the rst components o Split p0 \ (A UL A DR ) we obtain 2 jsplitp 0 \(A ULA DR )j dierent values or jja HIGH jj (including 0 and ), and by varying the second components we obtain 2 jsplitp 0 \(A ULA DR )j dierent values or jja LOW jj (also including 0 and ). Rows which correspond to the assignment jja HIGH jj = 0 and columns which correspond to the assignment jja LOW jj = 0 are removed. 4
15 The remaining submatrix M has 2 jsplitp 0 \(A ULA DR )j? dierent rows corresponding to dierent values or jja HIGH jj and 2 jsplitp 0 \(A ULA DR )j? dierent columns corresponding to dierent values or jja LOW jj. The entry m ij corresponding to jja HIGH jj and jja LOW jj is given by m ij = jja HIGHjj 2 n 2?p 0+ + jja LOW jj 2 = jja n HIGHjj 2 2?p 0 + jja LOWjj 2 n 2?p n 2?p 0 + : For jja HIGH jj = jja LOW jj we have m ij = jja HIGH jj, or jja LOW jj < jja HIGH jj we have m ij = jja HIGH jj?. (m ij < jja HIGH jj? would imply jja HIGH jj 2 n 2?p jja LOW jj, which cannot be true.) I the rows with values jja HIGH jj are ordered with increasing values x = ; x 2 ; x 3 ; : : : and the columns with values jja LOW jj are also ordered with increasing values, we obtain the ollowing submatrix M = 0 x 2? x 2 x 3? x 3? x 3 x 4? x 4? x 4? x x max? x max? x max? x max and similar to the special case we see that this matrix M has ull ran. Since jsplit p0 \ (A UL A DR )j > n 6, the ran is at least 2 n 6? and thus the ran o the original communication matrix is also in (2 n 6 ). C A Using Theorems 3 and we nally obtain the ollowing corollary: Corollary mtbdds, evbdds, bmds, *bmds, hdds, *bmds and *phdds require representations o size (2 n 6 ) or division (regardless o the variable order). 5 Conclusions We proved an exponential lower bound on the size o word-level representations or integer dividers. The proo could be done \simultaneously" or all word-level DDs by the introduction o Word-Level Linear Combination Diagrams (wlcds) as a generic word-level DD. They turned out to be a powerul tool to characterize the limits o the word-level DD-concept. Concerning division our result gives the ollowing hints or uture wor: Since word-level DDs are not suitable as a data structure at least as long as they are used or the representation o the input-output behaviour, new methods have to be developed. I existing 5
16 DDs are still to be used, e.g. the structure o the circuit might be considered to chec whether a hierachical substitution based approach is easible. On the other hand, it is an interesting open question, which type o (DD-similar) data structure is powerul enough to allow polynomial representation o division and ecient manipulation or verication at the same time. Reerences [] R.I. Bahar, E.A. Frohm, C.M. Gaona, G.D. Hachtel, E. Macii, A. Pardo, and F. Somenzi. Algebraic decision diagrams and their application. In Int'l Con. on CAD, pages 88{9, 993. [2] B. Becer, R. Drechsler, and R. Enders. On the computational power o bit-level and word-level decision diagrams. In ASP Design Automation Con., pages 46{467, 997. [3] R.E. Bryant. Graph - based algorithms or Boolean unction manipulation. IEEE Trans. on Comp., 35(8):677{69, 986. [4] R.E. Bryant. On the complexity o VLSI implementations and graph representations o Boolean unctions with application to integer multiplication. IEEE Trans. on Comp., 40:205{23, 99. [5] R.E. Bryant. Symbolic Boolean manipulation with ordered binary decision diagrams. ACM, Comp. Surveys, 24:293{38, 992. [6] R.E. Bryant and Y.-A. Chen. Verication o arithmetic unctions with binary moment diagrams. In Design Automation Con., pages 535{54, 995. [7] Y.-A. Chen and R.E. Bryant. *PHDD: An ecient graph representation or oating point circuit verication. In Int'l Con. on CAD, pages 2{7, 997. [8] E.M. Clare, M. Fujita, and X. Zhao. Hybrid decision diagrams - overcoming the limitations o MTBDDs and BMDs. In Int'l Con. on CAD, pages 59{63, 995. [9] E.M. Clare, K.L. McMillan, X. Zhao, M. Fujita, and J. Yang. Spectral transorms or large Boolean unctions with application to technology mapping. In Design Automation Con., pages 54{60, 993. [0] R. Drechsler, B. Becer, and S. Ruppertz. K*BMDs: A new data structure or verication. In European Design & Test Con., pages 2{8, 996. [] R. Drechsler, A. Sarabi, M. Theobald, B. Becer, and M.A. Perowsi. Ecient representation and manipulation o switching unctions based on ordered Kronecer unctional decision diagrams. In Design Automation Con., pages 45{49, 994. [2] M. Fujita, H. Fujisawa, and N. Kawato. Evaluation and improvements o Boolean comparison method based on binary decision diagrams. In Int'l Con. on CAD, pages 2{5, 988. [3] U. Kebschull, E. Schubert, and W. Rosenstiel. Multilevel logic synthesis based on unctional decision diagrams. In European Con. on Design Automation, pages 43{47, 992. [4] Y.-T. Lai and S. Sastry. Edge-valued binary decision diagrams or multi-level hierarchical verication. In Design Automation Con., pages 608{63, 992. [5] S. Mali, A.R. Wang, R.K. Brayton, and A.L. Sangiovanni-Vincentelli. Logic verication using binary decision diagrams in a logic synthesis environment. In Int'l Con. on CAD, pages 6{9, 988. [6] M. Naanishi. An exponential lower bound on the size o a binary moment diagram representing division. Master's thesis, Osaa University, [7] S. Waac. On the Descriptive and Algorithmic Power o Parity Ordered Binary Decision Diagrams, volume 200 o LNCS. Symp. on Theoretical Aspects o Comp. Science,
Boolean decision diagrams and SAT-based representations
Boolean decision diagrams and SAT-based representations 4th July 200 So far we have seen Kripke Structures 2 Temporal logics (and their semantics over Kripke structures) 3 Model checking of these structures
More informationBDD Based Upon Shannon Expansion
Boolean Function Manipulation OBDD and more BDD Based Upon Shannon Expansion Notations f(x, x 2,, x n ) - n-input function, x i = or f xi=b (x,, x n ) = f(x,,x i-,b,x i+,,x n ), b= or Shannon Expansion
More information13th International Conference on Relational and Algebraic Methods in Computer Science (RAMiCS 13)
13th International Conference on Relational and Algebraic Methods in Computer Science (RAMiCS 13) Relation Algebras, Matrices, and Multi-Valued Decision Diagrams Francis Atampore and Dr. Michael Winter
More informationA Unifying Approach to Edge-valued and Arithmetic Transform Decision Diagrams
Automation and Remote Control, Vol. 63, No., 2002, pp. 25 38. Translated from Avtomatika i Telemekhanika, No., 2002, pp. 40 53. Original Russian Text Copyright c 2002 by Moraga, Sasao, Stanković. AUTOMATA
More informationPreprint from Workshop Notes, International Workshop on Logic Synthesis (IWLS 97), Tahoe City, California, May 19-21, 1997
Preprint from Workshop Notes, International Workshop on Logic Synthesis (IWLS 97), Tahoe City, California, May 19-21, 1997 Multi-output Functional Decomposition with Exploitation of Don't Cares Christoph
More informationLinear Algebra (part 1) : Vector Spaces (by Evan Dummit, 2017, v. 1.07) 1.1 The Formal Denition of a Vector Space
Linear Algebra (part 1) : Vector Spaces (by Evan Dummit, 2017, v. 1.07) Contents 1 Vector Spaces 1 1.1 The Formal Denition of a Vector Space.................................. 1 1.2 Subspaces...................................................
More informationSimpler Functions for Decompositions
Simpler Functions or Decompositions Bernd Steinbach Freiberg University o Mining and Technology, Institute o Computer Science, D-09596 Freiberg, Germany Abstract. This paper deals with the synthesis o
More informationMulti-Terminal Multi-Valued Decision Diagrams for Characteristic Function Representing Cluster Decomposition
22 IEEE 42nd International Symposium on Multiple-Valued Logic Multi-Terminal Multi-Valued Decision Diagrams for Characteristic Function Representing Cluster Decomposition Hiroki Nakahara, Tsutomu Sasao,
More informationTitle. Citation Information Processing Letters, 112(16): Issue Date Doc URLhttp://hdl.handle.net/2115/ Type.
Title Counterexamples to the long-standing conjectur Author(s) Yoshinaka, Ryo; Kawahara, Jun; Denzumi, Shuhei Citation Information Processing Letters, 112(16): 636-6 Issue Date 2012-08-31 Doc URLhttp://hdl.handle.net/2115/50105
More informationby Symbolic Spectral Analysis of Boolean Functions Politecnico di Torino Torino. ITALY 10129
Predicting the Functional Complexity of Combinational Circuits by Symbolic Spectral Analysis of Boolean Functions Enrico Macii Massimo Poncino Politecnico di Torino Dip. di Automatica e Informatica Torino.
More informationNotes on the Matrix-Tree theorem and Cayley s tree enumerator
Notes on the Matrix-Tree theorem and Cayley s tree enumerator 1 Cayley s tree enumerator Recall that the degree of a vertex in a tree (or in any graph) is the number of edges emanating from it We will
More informationPolynomial Methods for Component Matching and Verification
Polynomial Methods for Component Matching and Verification James Smith Stanford University Computer Systems Laboratory Stanford, CA 94305 1. Abstract Component reuse requires designers to determine whether
More informationLecture 9 : PPAD and the Complexity of Equilibrium Computation. 1 Complexity Class PPAD. 1.1 What does PPAD mean?
CS 599: Algorithmic Game Theory October 20, 2010 Lecture 9 : PPAD and the Complexity of Equilibrium Computation Prof. Xi Chen Scribes: Cheng Lu and Sasank Vijayan 1 Complexity Class PPAD 1.1 What does
More informationNew Hierarchies of Representations RM97, September MAIN CONTRIBUTIONS OF THIS PAPER Generalizations of the Generalized Kronecker representation
HIERARCHIES OF AND/EXOR NEW DECISION DIAGRAMS, LATTICE TREES, CANONICAL FORMS, AND DIAGRAMS, LAYOUTS REGULAR Hierarchies of Representations RM97, September 1997 1 New Marek Perkowski, Lech Jozwiak y, Rolf
More information1 Basic Combinatorics
1 Basic Combinatorics 1.1 Sets and sequences Sets. A set is an unordered collection of distinct objects. The objects are called elements of the set. We use braces to denote a set, for example, the set
More informationCoins with arbitrary weights. Abstract. Given a set of m coins out of a collection of coins of k unknown distinct weights, we wish to
Coins with arbitrary weights Noga Alon Dmitry N. Kozlov y Abstract Given a set of m coins out of a collection of coins of k unknown distinct weights, we wish to decide if all the m given coins have the
More information(C) The rationals and the reals as linearly ordered sets. Contents. 1 The characterizing results
(C) The rationals and the reals as linearly ordered sets We know that both Q and R are something special. When we think about about either o these we usually view it as a ield, or at least some kind o
More informationCounter Examples to the Conjecture on the Complexity of BDD Binary Operations
TCS -TR-A-11-52 TCS Technical Report Counter Examples to the Conjecture on the Complexity of BDD Binary Operations by Ryo Yoshinaka, Jun Kawahara, Shuhei Denzumi, Hiroki Arimura and Shin-ichi Minato Division
More informationRepresentations of All Solutions of Boolean Programming Problems
Representations of All Solutions of Boolean Programming Problems Utz-Uwe Haus and Carla Michini Institute for Operations Research Department of Mathematics ETH Zurich Rämistr. 101, 8092 Zürich, Switzerland
More informationMinimal enumerations of subsets of a nite set and the middle level problem
Discrete Applied Mathematics 114 (2001) 109 114 Minimal enumerations of subsets of a nite set and the middle level problem A.A. Evdokimov, A.L. Perezhogin 1 Sobolev Institute of Mathematics, Novosibirsk
More informationVector Space Basics. 1 Abstract Vector Spaces. 1. (commutativity of vector addition) u + v = v + u. 2. (associativity of vector addition)
Vector Space Basics (Remark: these notes are highly formal and may be a useful reference to some students however I am also posting Ray Heitmann's notes to Canvas for students interested in a direct computational
More information[3] R.M. Colomb and C.Y.C. Chung. Very fast decision table execution of propositional
- 14 - [3] R.M. Colomb and C.Y.C. Chung. Very fast decision table execution of propositional expert systems. Proceedings of the 8th National Conference on Articial Intelligence (AAAI-9), 199, 671{676.
More informationOlivier Coudert Jean Christophe Madre Henri Fraisse. Bull Corporate Research Center. Rue Jean Jaures Les Clayes-sous-bois, FRANCE
A New Viewpoint on Two-Level Logic Minimization Olivier Coudert Jean Christophe Madre Henri Fraisse Bull Corporate Research Center Rue Jean Jaures 78340 Les Clayes-sous-bois, FRANCE Abstract This paper
More informationGF(4) Based Synthesis of Quaternary Reversible/Quantum Logic Circuits
GF(4) ased Synthesis o Quaternary Reversible/Quantum Logic Circuits MOZAMMEL H. A. KHAN AND MAREK A. PERKOWSKI Department o Computer Science and Engineering, East West University 4 Mohakhali, Dhaka, ANGLADESH,
More informationBounds on the OBDD-Size of Integer Multiplication via Universal Hashing
Bounds on the OBDD-Size of Integer Multiplication via Universal Hashing Philipp Woelfel Dept. of Computer Science University Dortmund D-44221 Dortmund Germany phone: +49 231 755-2120 fax: +49 231 755-2047
More informationMath 42, Discrete Mathematics
c Fall 2018 last updated 12/05/2018 at 15:47:21 For use by students in this class only; all rights reserved. Note: some prose & some tables are taken directly from Kenneth R. Rosen, and Its Applications,
More information2. ETA EVALUATIONS USING WEBER FUNCTIONS. Introduction
. ETA EVALUATIONS USING WEBER FUNCTIONS Introduction So ar we have seen some o the methods or providing eta evaluations that appear in the literature and we have seen some o the interesting properties
More informationBasing Decisions on Sentences in Decision Diagrams
Proceedings of the Twenty-Sixth AAAI Conference on Artificial Intelligence Basing Decisions on Sentences in Decision Diagrams Yexiang Xue Department of Computer Science Cornell University yexiang@cs.cornell.edu
More informationLecture 14 - P v.s. NP 1
CME 305: Discrete Mathematics and Algorithms Instructor: Professor Aaron Sidford (sidford@stanford.edu) February 27, 2018 Lecture 14 - P v.s. NP 1 In this lecture we start Unit 3 on NP-hardness and approximation
More informationA Lower Bound Technique for Nondeterministic Graph-Driven Read-Once-Branching Programs and its Applications
A Lower Bound Technique for Nondeterministic Graph-Driven Read-Once-Branching Programs and its Applications Beate Bollig and Philipp Woelfel FB Informatik, LS2, Univ. Dortmund, 44221 Dortmund, Germany
More informationDivisor matrices and magic sequences
Discrete Mathematics 250 (2002) 125 135 www.elsevier.com/locate/disc Divisor matrices and magic sequences R.H. Jeurissen Mathematical Institute, University of Nijmegen, Toernooiveld, 6525 ED Nijmegen,
More information1 Preliminaries We recall basic denitions. A deterministic branching program P for computing a Boolean function h n : f0; 1g n! f0; 1g is a directed a
Some Separation Problems on Randomized OBDDs Marek Karpinski Rustam Mubarakzjanov y Abstract We investigate the relationships between complexity classes of Boolean functions that are computable by polynomial
More informationUpper and Lower Bounds for Tree-like. Cutting Planes Proofs. Abstract. In this paper we study the complexity of Cutting Planes (CP) refutations, and
Upper and Lower Bounds for Tree-like Cutting Planes Proofs Russell Impagliazzo Toniann Pitassi y Alasdair Urquhart z UCSD UCSD and U. of Pittsburgh University of Toronto Abstract In this paper we study
More information1 Matrices and Systems of Linear Equations
Linear Algebra (part ) : Matrices and Systems of Linear Equations (by Evan Dummit, 207, v 260) Contents Matrices and Systems of Linear Equations Systems of Linear Equations Elimination, Matrix Formulation
More informationChapter 1. Comparison-Sorting and Selecting in. Totally Monotone Matrices. totally monotone matrices can be found in [4], [5], [9],
Chapter 1 Comparison-Sorting and Selecting in Totally Monotone Matrices Noga Alon Yossi Azar y Abstract An mn matrix A is called totally monotone if for all i 1 < i 2 and j 1 < j 2, A[i 1; j 1] > A[i 1;
More informationRivest-Vuillemin Conjecture Is True for Monotone Boolean Functions with Twelve Variables. Technical Report
Rivest-Vuillemin Conjecture Is True or Monotone Boolean Functions with Twelve Variables Technical Report Department o Computer Science and Engineering University o Minnesota 4-192 EECS Building 200 Union
More informationMin-Rank Conjecture for Log-Depth Circuits
Min-Rank Conjecture for Log-Depth Circuits Stasys Jukna a,,1, Georg Schnitger b,1 a Institute of Mathematics and Computer Science, Akademijos 4, LT-80663 Vilnius, Lithuania b University of Frankfurt, Institut
More informationx 1 x 2 x 3 x 4 x 5 x 6 x 7 x 8 x 9 x 10 x 11 x 12 x 13 x 14 x 15 x 16
DECOMPOS: An Integrated System for Functional Decomposition Tsutomu Sasao and Munehiro Matsuura Department of Computer Science and Electronics Kyushu Institute of Technology Iizuka 820-8502, Japan Abstract
More information16 Chapter 3. Separation Properties, Principal Pivot Transforms, Classes... for all j 2 J is said to be a subcomplementary vector of variables for (3.
Chapter 3 SEPARATION PROPERTIES, PRINCIPAL PIVOT TRANSFORMS, CLASSES OF MATRICES In this chapter we present the basic mathematical results on the LCP. Many of these results are used in later chapters to
More informationmatic scaling, ii) it can provide or bilateral power amplication / attenuation; iii) it ensures the passivity o the closed loop system with respect to
Passive Control o Bilateral Teleoperated Manipulators Perry Y. Li Department o Mechanical Engineering University o Minnesota 111 Church St. SE Minneapolis MN 55455 pli@me.umn.edu Abstract The control o
More informationContents. 2.1 Vectors in R n. Linear Algebra (part 2) : Vector Spaces (by Evan Dummit, 2017, v. 2.50) 2 Vector Spaces
Linear Algebra (part 2) : Vector Spaces (by Evan Dummit, 2017, v 250) Contents 2 Vector Spaces 1 21 Vectors in R n 1 22 The Formal Denition of a Vector Space 4 23 Subspaces 6 24 Linear Combinations and
More informationFraction-free Row Reduction of Matrices of Skew Polynomials
Fraction-free Row Reduction of Matrices of Skew Polynomials Bernhard Beckermann Laboratoire d Analyse Numérique et d Optimisation Université des Sciences et Technologies de Lille France bbecker@ano.univ-lille1.fr
More information2 Generating Functions
2 Generating Functions In this part of the course, we re going to introduce algebraic methods for counting and proving combinatorial identities. This is often greatly advantageous over the method of finding
More information1 Introduction In [2] a method for computing the Walsh spectrum in R-encoding of a completely specified binary-valued function was presented based on
Computation of Discrete Function Chrestenson Spectrum Using Cayley Color Graphs Λ Mitchell A. Thornton Southern Methodist University Dallas, Texas mitch@engr.smu.edu D. Michael Miller University of Victoria
More informationExtracted from a working draft of Goldreich s FOUNDATIONS OF CRYPTOGRAPHY. See copyright notice.
106 CHAPTER 3. PSEUDORANDOM GENERATORS Using the ideas presented in the proofs of Propositions 3.5.3 and 3.5.9, one can show that if the n 3 -bit to l(n 3 ) + 1-bit function used in Construction 3.5.2
More informationoutput H = 2*H+P H=2*(H-P)
Ecient Algorithms for Multiplication on Elliptic Curves by Volker Muller TI-9/97 22. April 997 Institut fur theoretische Informatik Ecient Algorithms for Multiplication on Elliptic Curves Volker Muller
More informationPreliminaries and Complexity Theory
Preliminaries and Complexity Theory Oleksandr Romanko CAS 746 - Advanced Topics in Combinatorial Optimization McMaster University, January 16, 2006 Introduction Book structure: 2 Part I Linear Algebra
More informationNumerical Methods - Lecture 2. Numerical Methods. Lecture 2. Analysis of errors in numerical methods
Numerical Methods - Lecture 1 Numerical Methods Lecture. Analysis o errors in numerical methods Numerical Methods - Lecture Why represent numbers in loating point ormat? Eample 1. How a number 56.78 can
More informationchapter 12 MORE MATRIX ALGEBRA 12.1 Systems of Linear Equations GOALS
chapter MORE MATRIX ALGEBRA GOALS In Chapter we studied matrix operations and the algebra of sets and logic. We also made note of the strong resemblance of matrix algebra to elementary algebra. The reader
More informationSolving Continuous Linear Least-Squares Problems by Iterated Projection
Solving Continuous Linear Least-Squares Problems by Iterated Projection by Ral Juengling Department o Computer Science, Portland State University PO Box 75 Portland, OR 977 USA Email: juenglin@cs.pdx.edu
More informationOn the Girth of (3,L) Quasi-Cyclic LDPC Codes based on Complete Protographs
On the Girth o (3,L) Quasi-Cyclic LDPC Codes based on Complete Protographs Sudarsan V S Ranganathan, Dariush Divsalar and Richard D Wesel Department o Electrical Engineering, University o Caliornia, Los
More informationDeterminants of Partition Matrices
journal of number theory 56, 283297 (1996) article no. 0018 Determinants of Partition Matrices Georg Martin Reinhart Wellesley College Communicated by A. Hildebrand Received February 14, 1994; revised
More informationChapter 6. Computational Complexity of Complementary Pivot Methods 301 ~q(n) =(2 n 2 n;1 ::: 2) T ^q(n) =(;2 n ;2 n ; 2 n;1 ;2 n ; 2 n;1 ; 2 n;2 ::: ;
Chapter 6 COMPUTATIONAL COMPLEXITY OF COMPLEMENTARY PIVOT METHODS In this Chapter, we discuss the worst case behavior of the computational growth requirements of the complementary and principal pivot methods
More informationLie Groups for 2D and 3D Transformations
Lie Groups for 2D and 3D Transformations Ethan Eade Updated May 20, 2017 * 1 Introduction This document derives useful formulae for working with the Lie groups that represent transformations in 2D and
More informationON COST MATRICES WITH TWO AND THREE DISTINCT VALUES OF HAMILTONIAN PATHS AND CYCLES
ON COST MATRICES WITH TWO AND THREE DISTINCT VALUES OF HAMILTONIAN PATHS AND CYCLES SANTOSH N. KABADI AND ABRAHAM P. PUNNEN Abstract. Polynomially testable characterization of cost matrices associated
More informationRandomized Complexity Lower Bounds. D. Grigoriev 1. The Pennsylvania State University. University Park, PA
Randomized Complexity Lower Bounds D. Grigoriev Departments o Mathematics and Computer Science The Pennsylvania State University University Par, PA 6802 dima@cse.psu.edu The complexity lower bound (log
More information3.1 Basic properties of real numbers - continuation Inmum and supremum of a set of real numbers
Chapter 3 Real numbers The notion of real number was introduced in section 1.3 where the axiomatic denition of the set of all real numbers was done and some basic properties of the set of all real numbers
More informationFeedback Linearization Lectures delivered at IIT-Kanpur, TEQIP program, September 2016.
Feedback Linearization Lectures delivered at IIT-Kanpur, TEQIP program, September 216 Ravi N Banavar banavar@iitbacin September 24, 216 These notes are based on my readings o the two books Nonlinear Control
More informationSymbolic Model Checking with ROBDDs
Symbolic Model Checking with ROBDDs Lecture #13 of Advanced Model Checking Joost-Pieter Katoen Lehrstuhl 2: Software Modeling & Verification E-mail: katoen@cs.rwth-aachen.de December 14, 2016 c JPK Symbolic
More informationMinimum Repair Bandwidth for Exact Regeneration in Distributed Storage
1 Minimum Repair andwidth for Exact Regeneration in Distributed Storage Vivec R Cadambe, Syed A Jafar, Hamed Malei Electrical Engineering and Computer Science University of California Irvine, Irvine, California,
More informationFinite Dimensional Hilbert Spaces are Complete for Dagger Compact Closed Categories (Extended Abstract)
Electronic Notes in Theoretical Computer Science 270 (1) (2011) 113 119 www.elsevier.com/locate/entcs Finite Dimensional Hilbert Spaces are Complete or Dagger Compact Closed Categories (Extended bstract)
More informationThe complexity of SPP formula minimization
The complexity of SPP formula minimization David Buchfuhrer Computer Science Department California Institute of Technology Pasadena, CA 91125 dave@cs.caltech.edu Abstract Circuit minimization is a useful
More informationReduced Ordered Binary Decision Diagrams
Reduced Ordered Binary Decision Diagrams Lecture #13 of Advanced Model Checking Joost-Pieter Katoen Lehrstuhl 2: Software Modeling & Verification E-mail: katoen@cs.rwth-aachen.de June 5, 2012 c JPK Switching
More information1 Introduction It will be convenient to use the inx operators a b and a b to stand for maximum (least upper bound) and minimum (greatest lower bound)
Cycle times and xed points of min-max functions Jeremy Gunawardena, Department of Computer Science, Stanford University, Stanford, CA 94305, USA. jeremy@cs.stanford.edu October 11, 1993 to appear in the
More informationLecture : Feedback Linearization
ecture : Feedbac inearization Niola Misovic, dipl ing and Pro Zoran Vuic June 29 Summary: This document ollows the lectures on eedbac linearization tought at the University o Zagreb, Faculty o Electrical
More informationLinear Algebra (part 1) : Matrices and Systems of Linear Equations (by Evan Dummit, 2016, v. 2.02)
Linear Algebra (part ) : Matrices and Systems of Linear Equations (by Evan Dummit, 206, v 202) Contents 2 Matrices and Systems of Linear Equations 2 Systems of Linear Equations 2 Elimination, Matrix Formulation
More informationBinary decision diagrams for security protocols
for Instytut Informatyki Teoretycznej i Stosowanej Politechnika Częstochowska 4 czerwca 2012 roku 1 2 3 4 Infrastructure with Intruder Threat template 5 References BDD definition Definition An BDD G
More informationLinear Algebra, 4th day, Thursday 7/1/04 REU Info:
Linear Algebra, 4th day, Thursday 7/1/04 REU 004. Info http//people.cs.uchicago.edu/laci/reu04. Instructor Laszlo Babai Scribe Nick Gurski 1 Linear maps We shall study the notion of maps between vector
More informationObjectives. By the time the student is finished with this section of the workbook, he/she should be able
FUNCTIONS Quadratic Functions......8 Absolute Value Functions.....48 Translations o Functions..57 Radical Functions...61 Eponential Functions...7 Logarithmic Functions......8 Cubic Functions......91 Piece-Wise
More informationRandom Benchmark Circuits with Controlled Attributes. [iwama,
Random Benchmark Circuits with Controlled Attributes Kazuo Iwama 3 Kensuke Hino y Hiroyuki Kurokawa z Sunao Sawada 3 3Dept of Computer Science and Communication Engineering Kyushu University, Japan yindustrial
More informationInteger Circuit Evaluation is PSPACE-complete. Ke Yang. Computer Science Department, Carnegie Mellon University, 5000 Forbes Ave.
Integer Circuit Evaluation is PSPACE-complete Ke Yang Computer Science Department, Carnegie Mellon University, 5000 Forbes Ave., Pittsburgh, PA 15213, USA E-mail: yangke@cmu.edu Key Words: PSPACE, Integer
More informationMTH 309 Supplemental Lecture Notes Based on Robert Messer, Linear Algebra Gateway to Mathematics
MTH 309 Supplemental Lecture Notes Based on Robert Messer, Linear Algebra Gateway to Mathematics Ulrich Meierfrankenfeld Department of Mathematics Michigan State University East Lansing MI 48824 meier@math.msu.edu
More informationLinear Regression and Its Applications
Linear Regression and Its Applications Predrag Radivojac October 13, 2014 Given a data set D = {(x i, y i )} n the objective is to learn the relationship between features and the target. We usually start
More informationare the q-versions of n, n! and . The falling factorial is (x) k = x(x 1)(x 2)... (x k + 1).
Lecture A jacques@ucsd.edu Notation: N, R, Z, F, C naturals, reals, integers, a field, complex numbers. p(n), S n,, b(n), s n, partition numbers, Stirling of the second ind, Bell numbers, Stirling of the
More informationAlgorithmic Approach to Counting of Certain Types m-ary Partitions
Algorithmic Approach to Counting of Certain Types m-ary Partitions Valentin P. Bakoev Abstract Partitions of integers of the type m n as a sum of powers of m (the so called m-ary partitions) and their
More informationLower Bounds for Cutting Planes Proofs. with Small Coecients. Abstract. We consider small-weight Cutting Planes (CP ) proofs; that is,
Lower Bounds for Cutting Planes Proofs with Small Coecients Maria Bonet y Toniann Pitassi z Ran Raz x Abstract We consider small-weight Cutting Planes (CP ) proofs; that is, Cutting Planes (CP ) proofs
More information1. Berichterstatter: Prof. Dr. Ch. Meinel (Universitat Trier) 2. Berichterstatter: Prof. Dr. B. Becker (Albert-Ludwigs-Universitat Freiburg) Datum der
Transformation Techniques for Decision Diagrams in Computer-Aided Design Dissertation zur Erlangung des Doktorgrades der Naturwissenschaften vorgelegt beim Fachbereich IV der Universitat Trier von Thorsten
More informationAN ALGEBRAIC INTERPRETATION OF THE q-binomial COEFFICIENTS. Michael Braun
International Electronic Journal of Algebra Volume 6 (2009) 23-30 AN ALGEBRAIC INTERPRETATION OF THE -BINOMIAL COEFFICIENTS Michael Braun Received: 28 February 2008; Revised: 2 March 2009 Communicated
More informationBinary Decision Diagrams
Binary Decision Diagrams Sungho Kang Yonsei University Outline Representing Logic Function Design Considerations for a BDD package Algorithms 2 Why BDDs BDDs are Canonical (each Boolean function has its
More informationLecture 16: Communication Complexity
CSE 531: Computational Complexity I Winter 2016 Lecture 16: Communication Complexity Mar 2, 2016 Lecturer: Paul Beame Scribe: Paul Beame 1 Communication Complexity In (two-party) communication complexity
More informationa (b + c) = a b + a c
Chapter 1 Vector spaces In the Linear Algebra I module, we encountered two kinds of vector space, namely real and complex. The real numbers and the complex numbers are both examples of an algebraic structure
More informationTechnische Universität München, Zentrum Mathematik Lehrstuhl für Angewandte Geometrie und Diskrete Mathematik. Combinatorial Optimization (MA 4502)
Technische Universität München, Zentrum Mathematik Lehrstuhl für Angewandte Geometrie und Diskrete Mathematik Combinatorial Optimization (MA 4502) Dr. Michael Ritter Problem Sheet 1 Homework Problems Exercise
More informationNumerical Solution of Ordinary Differential Equations in Fluctuationlessness Theorem Perspective
Numerical Solution o Ordinary Dierential Equations in Fluctuationlessness Theorem Perspective NEJLA ALTAY Bahçeşehir University Faculty o Arts and Sciences Beşiktaş, İstanbul TÜRKİYE TURKEY METİN DEMİRALP
More informationThe Nonapproximability of OBDD Minimization
The Nonapproximability of OBDD Minimization Detlef Sieling FB Informatik, LS II, Univ. Dortmund 44221 Dortmund, Fed. Rep. of Germany sieling@ls2.cs.uni-dortmund.de Abstract The size of Ordered Binary Decision
More informationContainment restrictions
Containment restrictions Tibor Szabó Extremal Combinatorics, FU Berlin, WiSe 207 8 In this chapter we switch from studying constraints on the set operation intersection, to constraints on the set relation
More informationBoolean Matching Using Generalized Reed-Muller Forms
oolean Matching Using Generalized Reed-Muller Forms Chien-Chung Tsai Malgorzata Marek-Sadowska Department of Electrical and Computer Engineering University of California Santa arbara, CA 93106 USA Abstract
More informationOn the Sizes of Decision Diagrams Representing the Set of All Parse Trees of a Context-free Grammar
Proceedings of Machine Learning Research vol 73:153-164, 2017 AMBN 2017 On the Sizes of Decision Diagrams Representing the Set of All Parse Trees of a Context-free Grammar Kei Amii Kyoto University Kyoto
More informationBinary Decision Diagrams
Binary Decision Diagrams Logic Circuits Design Seminars WS2010/2011, Lecture 2 Ing. Petr Fišer, Ph.D. Department of Digital Design Faculty of Information Technology Czech Technical University in Prague
More informationThe Cyclic Decomposition of a Nilpotent Operator
The Cyclic Decomposition of a Nilpotent Operator 1 Introduction. J.H. Shapiro Suppose T is a linear transformation on a vector space V. Recall Exercise #3 of Chapter 8 of our text, which we restate here
More informationThe cocycle lattice of binary matroids
Published in: Europ. J. Comb. 14 (1993), 241 250. The cocycle lattice of binary matroids László Lovász Eötvös University, Budapest, Hungary, H-1088 Princeton University, Princeton, NJ 08544 Ákos Seress*
More information2 EBERHARD BECKER ET AL. has a real root. Thus our problem can be reduced to the problem of deciding whether or not a polynomial in one more variable
Deciding positivity of real polynomials Eberhard Becker, Victoria Powers, and Thorsten Wormann Abstract. We describe an algorithm for deciding whether or not a real polynomial is positive semidenite. The
More informationParity Versions of 2-Connectedness
Parity Versions of 2-Connectedness C. Little Institute of Fundamental Sciences Massey University Palmerston North, New Zealand c.little@massey.ac.nz A. Vince Department of Mathematics University of Florida
More informationQuantum Multiple-Valued Decision Diagrams Containing Skipped Variables
Quantum Multiple-Valued Decision Diagrams Containing Skipped Variables DAVID Y. FEINSTEIN 1, MITCHELL A. THORNTON 1 Innoventions, Inc., 1045 Bissonnet Street, Houston, TX, USA Dept. of Computer Science
More informationExercises - Solutions
Chapter 1 Exercises - Solutions Exercise 1.1. The first two definitions are equivalent, since we follow in both cases the unique path leading from v to a sink and using only a i -edges leaving x i -nodes.
More informationREMARKS ON THE NUMBER OF LOGIC NETWORKS WITH SAME COMPLEXITY DERIVED FROM SPECTRAL TRANSFORM DECISION DIAGRAMS
REMARKS ON THE NUMBER OF LOGIC NETORKS ITH SAME COMPLEXITY DERIVED FROM SPECTRAL TRANSFORM DECISION DIAGRAMS Radomir S. Stanković Mark G. Karpovsky 1 Dept. of Computer Science, Faculty of Electronics,
More informationCOMPRESSED STATE SPACE REPRESENTATIONS - BINARY DECISION DIAGRAMS
QUALITATIVE ANALYIS METHODS, OVERVIEW NET REDUCTION STRUCTURAL PROPERTIES COMPRESSED STATE SPACE REPRESENTATIONS - BINARY DECISION DIAGRAMS LINEAR PROGRAMMING place / transition invariants state equation
More informationComputation of Floating Mode Delay in Combinational Circuits: Theory and Algorithms. Kurt Keutzer. Synopsys. Abstract
Computation of Floating Mode Delay in Combinational Circuits: Theory and Algorithms Srinivas Devadas MIT Cambridge, MA Kurt Keutzer Synopsys Mountain View, CA Sharad Malik Princeton University Princeton,
More informationAlgebraic Methods in Combinatorics
Algebraic Methods in Combinatorics Po-Shen Loh 27 June 2008 1 Warm-up 1. (A result of Bourbaki on finite geometries, from Răzvan) Let X be a finite set, and let F be a family of distinct proper subsets
More informationA Lower Bound Technique for Restricted Branching Programs and Applications
A Lower Bound Technique for Restricted Branching Programs and Applications (Extended Abstract) Philipp Woelfel FB Informatik, LS2, Univ. Dortmund, 44221 Dortmund, Germany woelfel@ls2.cs.uni-dortmund.de
More informationMitsuru Matsui , Ofuna, Kamakura, Kanagawa, 247, Japan. which are block ciphers with a 128-bit key, a 64-bit block and a variable
New Block Encryption Algorithm MISTY Mitsuru Matsui Inormation Technology R&D Center Mitsubishi Electric Corporation 5-1-1, Ouna, Kamakura, Kanagawa, 247, Japan matsui@iss.isl.melco.co.jp Abstract. We
More information