On integral generalized policy iteration for continuous-time linear quadratic regulations

This paper mathematically analyzes the integral generalized policy iteration (I-GPI) algorithms applied to a class of continuous-time linear quadratic regulation (LQR) problems with the unknown system matrix A$A$. GPI is the general idea of interacting policy evaluation and policy improvement steps of policy iteration (PI), for computing the optimal policy. We first introduce the update horizon ?$?$, and then show that (i) all of the I-GPI methods with the same ?$?$ can be considered equivalent and that (ii) the value function approximated in the policy evaluation step monotonically converges to the exact one as ?→∞$?\to \infty$. This reveals the relation between the computational complexity and the update (or time) horizon of I-GPI as well as between I-PI and I-GPI in the limit ?→∞$?\to \infty$. We also provide and discuss two modes of convergence of I-GPI; I-GPI behaves like PI in one mode, and in the other mode, it performs like value iteration for discrete-time LQR and infinitesimal GPI (?→0$?\to 0$). From these results, a new classification of the integral reinforcement learning is formed with respect to ?$?$. Two matrix inequality conditions for stability, the region of local monotone convergence, and data-driven (adaptive) implementation methods are also provided with detailed discussion. Numerical simulations are carried out for verification and further investigations.     

A polynomial-time algorithm for computing K-terminal residual reliability of d-trapezoid graphs

Consider a probabilistic graph G   in which the edges of E(G)$E(G)$ are perfectly reliable, but the vertices of V(G)$V(G)$ may fail with some known probabilities. Given a subset K   of V(G)$V(G)$, the K-terminal residual reliability of G is the probability that all operational vertices in K are connected to each other. This problem can be considered to be a generalization of two well-known reliability problems – the K-terminal reliability problem and the residual connectedness reliability problem.     

Algorithms for near solutions to polynomial equations

Let F(x,y)$F(x,y)$ be a polynomial over a field K$K$ and m$m$ a nonnegative integer. We call a polynomial g$g$ over K$K$ an m$m$-near solution of F(x,y)$F(x,y)$ if there exists a c∈K$c\in K$ such that F(x,g)=cx^m$F(x,g)=c{x}^m$, and the number c$c$ is called an m$m$-value of F(x,y)$F(x,y)$ corresponding to g$g$. In particular, c$c$ can be 0. Hence, by viewing F(x,y)=0$F(x,y)=0$ as a polynomial equation over K[x]$K\left[x\right]$ with variable y$y$, every solution of the equation F(x,y)=0$F(x,y)=0$ in K[x]$K\left[x\right]$ is also an m$m$-near solution. We provide an algorithm that gives all m$m$-near solutions of a given polynomial F(x,y)$F(x,y)$ over K$K$, and this algorithm is polynomial time reducible to solving one variable equations over K$K$. We introduce approximate solutions to analyze the algorithm. We also give some interesting properties of approximate solutions.     

Divergence bounded computable real numbers

A real x$x$ is called h$h$-bounded computable  , for some function h:N→N$h:\mathbb{N}\to \mathbb{N}$, if there is a computable sequence (_xs)$(x_s)$ of rational numbers which converges to x$x$ such that, for any n∈N$n\in \mathbb{N}$, at most h(n)$h(n)$ non-overlapping pairs of its members are separated by a distance larger than 2^-n$2^{-n}$. In this paper we discuss properties of h$h$-bounded computable reals for various functions h$h$. We will show a simple sufficient condition for a class of functions h$h$ such that the corresponding h$h$-bounded computable reals form an algebraic field. A hierarchy theorem for h$h$-bounded computable reals is also shown. Besides we compare semi-computability and weak computability with the h$h$-bounded computability for special functions h$h$.     

Fraction-free algorithm for the computation of diagonal forms matrices over Ore domains using Gröbner bases

This paper is a sequel to “Computing diagonal form and Jacobson normal form of a matrix using Gröbner bases” (Levandovskyy and Schindelar, 2011). We present a new fraction-free algorithm for the computation of a diagonal form of a matrix over a certain non-commutative Euclidean domain over a computable field with the help of Gröbner bases. This algorithm is formulated in a general constructive framework of non-commutative Ore localizations of G$G$-algebras (OLGAs). We use the splitting of the computation of a normal form for matrices over Ore localizations into the diagonalization and the normalization processes. Both of them can be made fraction-free. For a given matrix M$M$ over an OLGA R$R$, we provide a diagonalization algorithm to compute U,V$U,V$ and D$D$ with fraction-free entries such that UMV=D$UMV=D$ holds and D$D$ is diagonal. The fraction-free approach allows to obtain more information on the associated system of linear functional equations and its solutions, than the classical setup of an operator algebra with coefficients in rational functions. In particular, one can handle distributional solutions together with, say, meromorphic ones. We investigate Ore localizations of common operator algebras over K[x]$K\left[x\right]$ and use them in the unimodularity analysis of transformation matrices U,V$U,V$. In turn, this allows to lift the isomorphism of modules over an OLGA Euclidean domain to a smaller polynomial subring of it. We discuss the relation of this lifting with the solutions of the original system of equations. Moreover, we prove some new results concerning normal forms of matrices over non-simple domains. Our implementation in the computer algebra system Singular:Plural follows the fraction-free strategy and shows impressive performance, compared with methods which directly use fractions. In particular, we experience a moderate swell of coefficients and obtain simple transformation matrices. Thus the method we propose is well suited for solving nontrivial practical problems.     

Computing the homology of groups: The geometric way

In this paper, we present several algorithms related with the computation of the homology of groups, from a geometric perspective (that is to say, carrying out the calculations by means of simplicial sets and using techniques of Algebraic Topology). More concretely, we have developed some algorithms which, making use of the effective homology   method, construct the homology groups of Eilenberg–MacLane spaces K(G,1)$K(G,1)$ for different groups G$G$, allowing one in particular to determine the homology groups of G$G$.     

Approximately optimal trees for group key management with batch updates

We investigate the group key management problem for broadcasting applications. Previous work showed that, in handling key updates, batch rekeying can be more cost effective than individual rekeying. One model for batch rekeying is to assume that every user has probability p$p$ of being replaced by a new user during a batch period with the total number of users unchanged. Under this model, it was recently shown that an optimal key tree can be constructed in linear time when p$p$ is a constant and in O(n⁴)$O\left(n^4\right)$ time when p→0$p\to 0$. In this paper, we investigate more efficient algorithms for the case p→0$p\to 0$, i.e., when membership changes are sparse. We design an O(n)$O\left(n\right)$ heuristic algorithm for the sparse case and show that it produces a nearly 2-approximation to the optimal key tree. Simulation results show that its performance is even better in practice. We also design a refined heuristic algorithm and show that it achieves an approximation ratio of 1+?$1+?$ for any fixed ?>0$?>0$ and n$n$, as p→0$p\to 0$. Finally, we give another approximation algorithm for any p∈(0,0.693)$p\in (0,0.693)$ which is shown to be quite good by our simulations.     

Bounds on the bisection width for random d -regular graphs

In this paper we provide an explicit way to compute asymptotically almost sure upper bounds on the bisection width of random d$d$-regular graphs, for any value of d$d$. The upper bounds are obtained from the analysis of the performance of a randomized greedy algorithm to find bisections of d$d$-regular graphs. We provide bounds for 5≤d≤12$5\le d\le 12$. We also give empirical values of the size of the bisection found by the algorithm for some small values of d$d$ and compare them with numerical approximations of our theoretical bounds. Our analysis also gives asymptotic lower bounds for the size of the maximum bisection.     

On mimicking networks representing minimum terminal cuts

Given a capacitated undirected graph G=(V,E)$G=(V,E)$ with a set of terminals K⊂V$K\subset V$, a mimicking network   is a smaller graph H=(_VH,_EH)$H=(V_H,E_H)$ which contains the set of terminals K   and for every bipartition [U,K−U]$[U,K-U]$ of the terminals, the cost of the minimum cut separating U   from K−U$K-U$ in G is exactly equal to the cost of the minimum cut separating U   from K−U$K-U$ in H.     

An augmented weighted Tchebycheff method with adaptively chosen parameters for discrete bicriteria optimization problems

The augmented weighted Tchebycheff norm was introduced in the context of multicriteria optimization by Steuer and Choo [21] in order to avoid the generation of weakly nondominated points. It augments a weighted _l∞-norm$l_{\infty }\mathrm{-}\mathrm{norm}$ with an l₁-term, multiplied by a “small” parameter ρ>0$\rho >0$. However, the appropriate selection of the parameter ρ$\rho$ remained an open question: A too small value of ρ$\rho$ may cause numerical difficulties, while a too large value of ρ$\rho$ may lead to the oversight of some nondominated points.     

Finite derivation type for Rees matrix semigroups

This paper introduces the topological finiteness condition finite derivation type   (FDT) on the class of semigroups. This notion is naturally extended from the monoid case. With this new concept we are able to prove that if a Rees matrix semigroup M[S;I,J;P]$M[S;I,J;P]$ has FDT then the semigroup S$S$ also has FDT. Given a monoid S$S$ and a finitely presented Rees matrix semigroup M[S;I,J;P]$M[S;I,J;P]$ we prove that if the ideal of S$S$ generated by the entries of P$P$ has FDT, then so does M[S;I,J;P]$M[S;I,J;P]$. In particular, we show that, for a finitely presented completely simple semigroup M$M$, the Rees matrix semigroup M=M[S;I,J;P]$M=M[S;I,J;P]$ has FDT if and only if the group S$S$ has FDT.     

OptQC: An optimized parallel quantum compiler

The software package Qcompiler (Chen and Wang 2013) provides a general quantum compilation framework, which maps any given unitary operation into a quantum circuit consisting of a sequential set of elementary quantum gates. In this paper, we present an extended software OptQC  , which finds permutation matrices P$P$ and Q$Q$ for a given unitary matrix U$U$ such that the number of gates in the quantum circuit of U=Q^TP^TU^′PQ$U=Q^T{P}^T{U}^{\prime }PQ$ is significantly reduced, where U^′$U^{\prime }$ is equivalent to U$U$ up to a permutation and the quantum circuit implementation of each matrix component is considered separately. We extend further this software package to make use of high-performance computers with a multiprocessor architecture using MPI. We demonstrate its effectiveness in reducing the total number of quantum gates required for various unitary operators.     

Non-reducible descriptions for conditional Kolmogorov complexity

Assume that a program p$p$ on input a$a$ outputs b$b$. We are looking for a shorter program q$q$ having the same property (q(a)=b$q\left(a\right)=b$). In addition, we want q$q$ to be simple conditional to p$p$ (this means that the conditional Kolmogorov complexity K(q|p)$K\left(q\right|p)$ is negligible). In the present paper, we prove that sometimes there is no such program q$q$, even in the case when the complexity of p$p$ is much bigger than K(b|a)$K\left(b\right|a)$. We give three different constructions that use the game approach, probabilistic arguments and algebraic arguments, respectively.     

Paths and trails in edge-colored graphs

This paper deals with the existence and search for properly edge-colored paths/trails between two, not necessarily distinct, vertices s$s$ and t$t$ in an edge-colored graph from an algorithmic perspective. First we show that several versions of the s−t$s-t$ path/trail problem have polynomial solutions including the shortest path/trail case. We give polynomial algorithms for finding a longest properly edge-colored path/trail between s$s$ and t$t$ for a particular class of graphs and characterize edge-colored graphs without properly edge-colored closed trails. Next, we prove that deciding whether there exist k$k$ pairwise vertex/edge disjoint properly edge-colored s−t$s-t$ paths/trails in a c$c$-edge-colored graph G^c$G^c$ is NP-complete even for k=2$k=2$ and c=Ω(n²)$c=\Omega \left(n^2\right)$, where n$n$ denotes the number of vertices in G^c$G^c$. Moreover, we prove that these problems remain NP-complete for c$c$-edge-colored graphs containing no properly edge-colored cycles and c=Ω(n)$c=\Omega \left(n\right)$. We obtain some approximation results for those maximization problems together with polynomial results for some particular classes of edge-colored graphs.     

The number of roots of a lacunary bivariate polynomial on a line

We prove that a polynomial f∈R[x,y]$f\in \mathbb{R}[x,y]$ with t$t$ non-zero terms, restricted to a real line y=ax+b$y=ax+b$, either has at most 6t−4$6t-4$ zeros or vanishes over the whole line. As a consequence, we derive an alternative algorithm for deciding whether a linear polynomial y−ax−b∈K[x,y]$y-ax-b\in K[x,y]$ divides a lacunary polynomial f∈K[x,y]$f\in K[x,y]$, where K$K$ is a real number field. The number of bit operations performed by the algorithm is polynomial in the number of non-zero terms of f$f$, in the logarithm of the degree of f$f$, in the degree of the extension K/Q$K/\mathbb{Q}$ and in the logarithmic height of a$a$, b$b$ and f$f$.     

Claw finding algorithms using quantum walk

The claw finding problem has been studied in terms of query complexity as one of the problems closely connected to cryptography. Given two functions, f$f$ and g$g$, with domain sizes N$N$ and M$M$(N≤M)$(N\le M)$, respectively, and the same range, the goal of the problem is to find x$x$ and y$y$ such that f(x)=g(y)$f\left(x\right)=g\left(y\right)$. This problem has been considered in both quantum and classical settings in terms of query complexity. This paper describes an optimal algorithm that uses quantum walk to solve this problem. Our algorithm can be slightly modified to solve the more general problem of finding a tuple consisting of elements in the two function domains that has a prespecified property. It can also be generalized to find a claw of k$k$ functions for any constant integer k>1$k>1$, where the domain sizes of the functions may be different.     

Gröbner bases with respect to several orderings and multivariable dimension polynomials

Let D=K[X]$D=K\left[X\right]$ be a ring of Ore polynomials over a field K$K$ and let a partition of the set of indeterminates into p$p$ disjoint subsets be fixed. Considering D$D$ as a filtered ring with the natural p$p$-dimensional filtration, we introduce a special type of reduction in a free D$D$-module and develop the corresponding Gröbner basis technique (in particular, we obtain a generalization of the Buchberger Algorithm). Using such a modification of the Gröbner basis method, we prove the existence of a Hilbert-type dimension polynomial in p$p$ variables associated with a finitely generated filtered D$D$-module, give a method of computation and describe invariants of such a polynomial. The results obtained are applied in differential algebra where the classical theorems on differential dimension polynomials are generalized to the case of differential structures with several basic sets of derivation operators.