Abstract Poster PDF
Abstract Poster PDF
Abstract Poster PDF
Poster Abstracts
1
A Speed-Up of the Hermite Reduction for
Rational Functions
The Hermite reduction decomposes a rational function f (x) into two rational
functions g(x) and h(x) such that
dg
f= + h,
dx
and that h has a square-free denominator. The reduction is a basis for many
algorithms in symbolic integration.
Let K be the coefficient field of rational functions. During the Hermite
reduction, the following two calculations frequently occur:
The first calculation requires polynomial division, and the second requires
the extended Euclidean algorithm (EEA). Instead of expanding a1 · · · am or
using division and EEA naively, we compute r, s, and t using the factors.
Importing this trick into the Hermite reduction algorithm (linear version)
in [1], we gain a significant speed-up when K is a field of rational functions
in other variables.
The poster outlines our idea, reports experimental results, and presents
an application in the construction of Zeilberger pairs for bivariate rational
functions.
2
References
[1] M. Bronstein. Symbolic Integration I: Transcedental Functions, second
edition, Springer-Verlag, 2005.
3
Efficient Characteristic Set Algorithms for
Equation Solving in Finite Fields
Email: [email protected]
Consider a Liénard system
ẍ + f (x)ẋ + g(x) = 0,
where f (x) and g(x) are polynomials. Let F 0 (x) = f (x), ẏ = −g(x), we can
rewrite it into a two-dimensional form
ẋ = y − F (x), ẏ = −g(x). (1.1)
4
Let H(i, j) denote the maximumal number of limit cycles, where i is
the degree of f and j the degree of g. In 1977, Lins, de Melo and Pugh
considered system (1.1) and proved that H(2,1)=1. They conjectured that
if i = 2n or 2n + 1, there could be no more than n limit cycles for system
(1.1). Coppel proved that H(1,2)=1. In 1996 and 1997, Dumortier and Li
proved that H(1,3)=1 and H(2,2)=1.
The global results for the maximal number of limit cycles for system
(1.1) are relatively few. In recent years many results have been obtained for
small amplitude limit cycles for system (1.1). We use Ĥ(i, j) to denote the
maximal number of small amplitude limit cycles, which can be bifurcated
within a small neighborhood of the origin for system (1.1).
As we all known, the first step for constructing small amplitude limit
cycles is to compute the Liapunov quantities (or focal values) for a system
and then to solve the principal unknown from the corresponding polynomial
of Liapunov quantities after triangularizing all of them. In this paper, we
shall deal with the triangularized polynomials of Liapunov quantities by an
algorithm of real root isolation without solving the principal unknowns and
then construct small amplitude limit cycles based on the independence of
these Liapunov quantities.
Based on an algorithm for computing focal values given by Chicone C.
and Jacobs M.[1] and an algorithm of real root isolation for multivariant
polynomials by Lu Z.Y., He B. and Luo Y. [4], an algorithmic construction
of small amplitude limit cycles for a class of Liénard systems is proposed.
The maximal number of small amplitude limit cycles for a class of Liénard
systems are obtained.
THEOREM 1. [2, 3, 4] Ĥ(7, 4) = 8, Ĥ(7, 5) = 9, Ĥ(8, 4) = 9, Ĥ(5, 5) =
6, Ĥ(6, 5) = 8, Ĥ(4, 6) = 7, Ĥ(5, 6) = 8, Ĥ(4, 7) = 8, Ĥ(4, 8) = 9, Ĥ(8, 5) =
10.
Furthermore, we have a result for some general value Ĥ(n, m) .
» ¼ » ¼
n+2 n+2
THEOREM 2. [4] n+1− ≤ Ĥ(n, m) ≤ n+m−1− .
m+1 m+1
References
[1] Chicone C. and Jacobs M., Bifurcation of critical periods for plane
vector fields, Tran. AMS, 1989, 433-486.
5
[2] He B. Lin S. Lu Z., The Construction of Small-amplitude Limit Cycles
for Liénard Systems Based on an Algorithm of Real Root Isolation,
preprint.
[3] Lin S. Lu Z., The Number of Small Amplitude Limit cycles for a
polynomial Liénard system with degrees (8,5), preprint.
[4] Lu Z.Y., He B., Luo Y. An algorithm of real root isolation for poly-
nomial systems, Sci. Press, 2004 (in Chinese).
6
A New Advance in Dense Packing of Equal Circles in a Circle
( In honor of Prof. Wen-Tsun Wu’s ninetieth birthday )
7
at ISSAC’08 both based on polynomial SOS, which do not always exist, into
a universal algorithm for all inputs via Artin’s theorem.
Furthermore, we scrutinize the all-important process of converting the
numerical SOS numerators and denominators produced by block semidefi-
nite programming into an exact rational identity. We improve on our own
Newton iteration-based high precision refinement algorithm by compressing
the initial Gram matrices and by deploying rational vector recovery aside
from orthogonal projection. We successfully demonstrate our algorithm on
1. various exceptional SOS problems with necessary polynomial denomina-
tors from the literature, on 2. very large (thousands of variables) SOS lower
bound certificates for Rump’s model problem (up to n = 18, factor degree
= 17) and on 3. a proof of the monotone column permanent (MCP) con-
jecture for dimension 4, which is sufficient to show that 4 polynomials are
nonnegative for all real values of the variables.
8
like Isabelle, Coq and Mizar. The formalization of machine proof to four
color theorem and the Kepler’s conjecture can also be considered as specic
examples of this category.
In this poster we present our joint work on formalized topology in Is-
abelle/HOL. We see propositions and proofs to them step by step with logical
calculation exe- cuted according to given proof methods and propositions.
Since the proofs allow no logical gaps, most of those proofs are tedious. In
addition, unless enough formalized background knowledge, the main proof
stream is disturbed by that of preliminary facts, and this makes a proof
complicated. Even though the proof is correct, if it is too hard to read then
it is not helpful for us. Therefore we have to give concise proofs.
We analyze structure of mechanized topology, and present some points
to make short mechanized proofs running with the aid of automated reason-
ing. In section 1 we present a part of mechanized topology to show how a
lemma in topology is formalized. In section 2 we discuss maturity of formal-
ization language. In human language we have several expressions for one
object and we can use those expressions representing the same object freely.
But, in formalized mathematics, it takes long steps to see one expression
is the same as another one, even if we can see apparently these have the
same contents. In section 3 we discuss cutting out preliminary properties.
Proving such preliminary properties in advance and put them in some le,
we can write clear proof. In section 4 we discuss key ideas and a total idea
of a proof. Some lemmas can be proved only after giving a key idea found
by a/some mathematician/s. If a proof to a lemma requires long step, it is
quite hard to prove the lemma automatically, even conning to a part of the
proof it is also very hard to prove it automatically. If we have a mechanism
to interpret key ideas and a total idea of a proof, then it will be useful to
make an automated reasoning system. In section 5 we explain the proof
given in section 1 and discuss possibility of automated reasoning system.
9
in k[x1 , · · · , xn ]. We can quickly obtain the primary decomposition of I
when I is normal position with respect to a variable. Otherwise, when
I is not normal with respect to every variable, a general method is that
introduce a new variable z and pick c = (c1 , · · · , cn ) ∈ k n , then let g =
z − c1 x1 − · · · − cn xn ∈ k[x1 , · · · , xn , z], J =< I, g >,we get the extension
ideal J of I. And J is normal position with respect to the variable z. But
the selection of c in the extension ideal of I is an important and difficult
problem in the primary decomposition of I. In this paper we discuss this
problem and give an efficient method to pick c.
2. main method.
Let K is the algebraic closed field of k,Z1 , Z2 , · · · , Zr+s ∈ K n are the
zeroes of the ideal I. Assume that
Here Zj 6= Zk , j, k = 1, 2, · · · , r + s. Let
Z11 c1 + Z12 c2 + · · · + Z1n cn = b1
.
..
Zr1 c1 + Zr2 c2 + · · · + Zrn cn = br (2)
..
.
Zr+s,1 c1 + Zr+s,2 c2 + · · · + Zr+s,n cn = br+s
We request only b1 , b2 , · · · , br+s are pairwise different.We regard Z1 , Z2 , · · · ,
Zr+s as r+s vectors in K n . Without loss of generality, we assume Z1 , Z2 , · · · , Zr
is a maximal linear independent subset of Z1 , Z2 , · · · , Zr+s , then Pr we know
that
Pr we only need select b 1 , · · · , b r pairwise different such that i=1 bi aji 6=
b a
i=1 i li =
6 b k , then b ,
1 2b , · · · , b r+s are pairwise different,here 1 ≤ j 6= l ≤
s, 1 ≤ k ≤ r. Solving equation (2), we shall obtain c.
3. main result.
We proved that we can pick the above b1 , b2 , · · · , br . When s = 0, we
only need pick pairwise different b1 , b2 , · · · , br and solve(2),then We can eas-
ily get the requisition c1 , c2 , · · · , cn . When the value of s is little,picking
10
b1 , b2 , · · · , br is also easy. Since r ≤ n, so the number of bi (1 ≤ i ≤ r) is no
more than ci (1 ≤ i ≤ n), some time is much less. So we pick bi is much easy
than select ci directly. So our method is more easy than Beck’s find out a
list of c = (c1 , · · · , cn ).
JPEG2000 and H.264 are the latest image and video coding standards re-
spectively. Digital cinema is a new kind of application for super high def-
inition video. The DCI (Digital Cinema Initiative) specification published
in 2005 has selected JPEG2000 instead of H.264 as the video coding stan-
dard for digital cinema. In this paper, based on JPEG2000, H.264 intra and
inter-frame coding, we compare the coding efficiency and subjective image
quality on multiple series of test sequences from low to super high resolu-
tions. The experiment results demonstrate some regularity for JPEG2000
and H.264 video coding, and reveal JPEG2000 is more suitable for digital
cinema.
Chaolu T
Shanghai Maritime University
Email: [email protected]
11
presented. This is a new application of differential form characteristic set al-
gorithm (differential form Wu’s method) in the field of differential equations.
Guozhao Wang
Zhejiang University, China
Email: [email protected]
Mingsheng Wang
Institute of Software, CAS, China
Email: mingsheng [email protected]
12
Let n be an integer with n ≥ 1. Let k be a field, R = k[z1 , . . . , zn ] be
the polynomial ring in variables z1 , . . . , zn over k, and Q(R) be its fraction
field. Rl×m denotes the free module of l × m matrices with entries in R. We
also write R1×m as Rm which is a free module of rank m over R.
Let F ∈ Rl×m be of full row rank, we denote the greatest common divisor
of all the i × i minors of F by di (F ), i = 1, . . . , l. We set d(F ) = dl (F ), and
ρ(F ) denotes the submodule generated by rows of F .
In order to state the research problems, some basic definitions are needed:
Definition 0.1 Let F ∈ Rl×m be of full row rank. Then F is said to be:
(i) zero left prime (ZLP) if all the l × l minors of F generate the unit
ideal R;
(ii) minor left prime (MLP) if all the l × l minors of F are relatively
prime, i.e., d(F ) is a nonzero constant;
(iii) factor left prime (FLP) if in any polynomial matrix factorization
F = F1 F2 in which F1 is a square matrix, F1 is necessarily a unimodular
matrix, i.e., det F1 is a nonzero constant in k.
Zero right prime (ZRP), and minor right prime (MRP) etc. can be
similarly defined for matrices F ∈ Rm×l with m ≥ l.
Notice that ZLP ⇒ MLP ⇒ FLP. When n ≥ 3, these concepts are
pairwise different, when n = 2, ZLP is not equivalent to MLP, but MLP is
the same as FLP, when n = 1 all three concepts coincide.
Definition 0.2 Let F ∈ Rl×m be of full row rank, and f be a divisor (not
necessarily the gcd) of d(F ), that is, ai = f bi , where a1 , . . . , aβ are all
the l × l-minors, and bi ∈ R, i = 1, . . . , β. We say that F has a matrix
factorization with respect to f if F can be factorized as
F = G 1 F1 (1)
13
A Bivariate Preprocessing Paradigm for
Buchberger-Möller Algorithm
For the last almost three decades, since the famous Buchberger-Möller (BM)
algorithm emerged, there has been wide interest in vanishing ideals of affine
points and associated interpolation polynomials. For the sake of reducing
the complexity of BM algorithm, we propose a preprocessing paradigm for
BM algorithm in bivariate cases. First, we will give an estimation of certain
aspect of the geometry of the points. Next, we will introduce our main al-
gorithm for finding a maximal lower subset of the original points and then
constructing associated Newton-type polynomials that are part of the out-
put of BM algorithm w.r.t. some rearrangement of the order of the points.
These results can be used by BM procedure directly hence simplify the com-
putation.
Email: [email protected]
14
throughout a specic sequence of model checking, is called an invariance to
the sequence. The invariance can dramatically avoid the checking repeat-
edly. In this paper, we rst construct a formal framework of dynamic model
checking, and then propose an invariance theory of dynamic model checking
based on an iterative design process of ow control oriented hardware systems
(FC-oriented HS for short) described by Moore machines. The FC-oriented
HS is a kind of sequen- tial circuit system stressing to handle data trans-
formation. Furthermore, we show that some non-trivial CTL properties are
preserved in the it- eration. To our best knowledge, this is the rst research
on invariance of dynamic model checking along the iterative hardware sys-
tems design process. This theory is an extension and a supplement of the
classical CTL case.
Key words:dynamic model checking, invariance, iterative design, ow con-
trol oriented hardware system.
An Algorithm for
Vector Valued Osculatory Rational Interpolation
Based on Groebner Bases of R-modules
In this paper we convert the task of seeking the weak solution (~a(X), b(X))
of vector valued osculatory rational interpolation into computing the Groeb-
ner bases of R-submodule M of the free module over the polynomial ring.
We compute the Groebner bases recursively so that the algorithm for ra-
tional interpolation is Newton-type. Complexity is measured in terms of
max{deg(~a), deg(b) + ξ}, where ξ is a given integer used in defining the term
order of the R-module.
Keywords: Groebner bases, modules, weak interpolations, vector valued
osculatory rational interpolation.
15
Balanced Dense Polynomial Multiplication on Multi-cores
Email: [email protected]
16
Keywords: mKdV-Burgers equation, approximate symmetry reduction,
series reduction.
Gui-Fang Zhang
Beijing Forestry University, China
Email: [email protected]
In this paper, the invariance principle and principle of duality are intro-
duced. The uniform definition of adjoint operators of general operators
has been presented. As the applications, some systems of PDEs have been
solved, including linear PDEs with variable coefficients and some kinds of
nonlinear PDEs.
17
Hybrid Divide-and-Conquer Method for
Solving Deficient Polynomial Systems
(In honor of Prof. Wen-Tsun Wu’s ninetieth birthday)
18
A Human Eye Location Algorithm Based on
the Quasi Binary Image
A method for eyes localization is described in this paper. Its based on twice
binary images and twice vertical projection. Firstly, the original image is
decomposed with 8-level of wavelet , and then by vertical and horizontal
projection, the main area for eyes is gotten. Secondly, the given image is
decomposed with SVD again and 35% of the max singular value is set to be
a new threshold and to reconstruct the original image for obtaining the quasi
binary image which embodies eyes better. At last, eyes area is intercepted
from less-noise quasi binary image by already known coordinates and finally,
accurate position of eyes are achieved by vertical projection once again. Ex-
perimental results indicate that the proposed method is easy, feasible, and
fairly precise for eye localization.
Xinchao Zhao
Beijing University of Posts and Telecommunications, China
Email: [email protected]
Goal: It’s obvious that there are NO exact solutions satisfying all the
equations. So our goal is to find a vector X ∈ F2n×1 which satisfies as many
equations as possible.
19
Method: Population heuristic particle swarm algorithm(PSA) is used.
As far as I know, it’s the first time to adapt the PSA to solve such discrete
equations in F2 .
Results: An improved random disturbed PSA is proposed for the equa-
tions solving problem. Four randomly generated equations are solved with
sizes A ∈ F2100×20 , A ∈ F2300×50 , A ∈ F2500×100 and A ∈ F21000×200 .
20