1 Explicit Explore or Exploit (E 3 ) Algorithm
|
|
- Malcolm Morgan
- 5 years ago
- Views:
Transcription
1 2.997 Decision-Making in Lage-Scale Systems Mach 3 MIT, Sping 2004 Handout #2 Lectue Note 9 Explicit Exploe o Exploit (E 3 ) Algoithm Last lectue, we studied the Q-leaning algoithm: [ ] Q t+ (x t, a t ) = Q t (x t, a t ) + β t g (x t ) + π min Q t (x t+, a ) Q t (x t, a t ). a t An impotant chaacteistic of Q-leaning is that it is a model-fee appoach to leaning an optimal policy in an MDP with unknown paametes. In othe wods, thee is explicit attempt to model o estimate costs and/o tansition pobabilities the value of each action is estimated diectly though the Q-facto. Anothe appoach to the same poblem is to estimate the MDP paametes fom the data and find a policy based on the estimated paametes. In this lectue, we will study one such algoithm the Explicit Exploe o Exploit (E 3 ) algoithm, poposed by Keans and Singh []. The main ideas fo E 3 ae as follows: we divide states in two sets: a N N C known states unknown states known states have been visited sufficiently many times to ensue that Pˆa(x, y), ĝ a (x) ae accuate with high pobabilities an unknown state is moved to N when it has been visited at least m times fo some numbe m We intoduce two MDPs Mˆ N and M N. The MDP Mˆ N is pesented in Fig.. Its main chaacteistic is that the unknown states fom the oiginal MDP ae meged into a ecuent state x 0 with cost g a (x 0 ) = g max, a. The othe MDP M N has the same stuctue as Mˆ N but the estimated tansition pobabilities and costs ae eplaced with thei tue values. We now intoduce the algoithm.. Algoithm We will fist conside a vesion of E 3 which assumes knowledge of J ; the assumption will be lifted late. The E 3 algoithm poceeds as follows.. Let N =. Pick abitay state x 0. Let k = If x k / N, pefom balanced wandeing: If x k N, then a k = action chosen fewest times at state x k
2 attempt exploitation: If the optimal policy π fo Mˆ N has Ĵ ˆ (x k ) J (x k ) + β M N 2, stop. Retun x k and π ˆMN attempt exploation: Follow policy ˆπ S0 fo T steps whee T = α. ˆ Figue : Makov Decision Pocess M n Theoem With pobability no less than δ, E 3 will stop afte a numbe of actions and computation time ( ) poly,, S,, g max δ δ and etun a state x and policy u such that J u (x) J (x) + δ..2 Main Points The main points used fo poving Theoem ae as follows: (i) Thee exists m that is polynomially bounded such that, if all states in N have been visited at least m ˆ times, then M N is sufficiently close to M N. (ii) Balanced wandeing can only happen finitely many times. (iii) (a) J u,mn (x) J u (x) (b) J u,mn J u,mn β with high pobability ˆ 2 (iv) If exploitation is not possible, then thee is an exploation policy that eaches an unknown state afte T tansitions with high pobability. To show the fist main point, we conside the following lemma. Lemma Suppose a state x has been visited at least m times with each action a A x having been executed at least m A x times. Then, if ( ) m = poly S,, T, g max,, log, va(g) δ δ 2
3 we have, w.p. δ, ( ( ) ) 2 Pˆa(x, y) P a (x, y) = O δ S g max ( ( ) ) 2 ĝ a (x) g a (x) = O δ S g max The poof of this lemma is a diectly application of the Chenoff bound, which states that, if z, z 2,... ae i.i.d. Benoulli andom vaiables, then n zi Ez n i= (SLLN) ( ) ( ) n nδ 2 P z i Ez > δ 2 exp n i= 2 The main point (ii) follows fom pigeonhole pinciple: afte (m ) S balanced wandeing steps, at least one state will have to become known The main point iii(a) follows fom the next lemma. Lemma 2 Fo all policy u, J u,mn (x) J u (x), x. Poof: Tivial fo x / N since J u,mn (x) = gmax α J u (x). If x N, take T = inf{t : x t / N}. Then [ T ] J π t g π t u (x) = E u(x t ) + gu(x t ) t=0 t=t [ ] T E π t g u (x t ) + π T gmax t=0 = J u,mn (x) To pove the main point iii(b), we fist intoduce the following definition. Definition Let M and M ˆ be two MDPs. Then M ˆ is a β-appoximation to M if Lemma 3 If T α log 2g max β( α) Pˆa(x, y) P a (x, y) ) ( and ˆM is an O δ β g a (x) ĝ a (x) β. α S g max J u,m J u,m δ. ˆ ) ) 2 appoximation of M, then, u, 3
4 Sketch of poof: Take a policy u and a stat state x. We conside paths of length T stating fom x: p = x 0, x, x 2,..., x T whee p denotes the path. Note that [ ] J u,m (x) = P u,m (p)g u (p) + E π t g u (x t ), p t=t + whee P u,m (p) = P u,m (x 0, x )P u,m (x, x 2 )... P u,m (x T, x T ) is the pobability of obseving path p and is the discounted cost associated with path p. By selecting T popely, we can have T gu(p) = π g u (x t ) [ ] π T g max E π t g u (x t ) δ t=t + Recall that Pa(x, y) Pˆa(x, y) β. We conside two kinds of paths: (a) paths containing at least one tansition x t, x t+ in the set R such that P u (x t, x t+ ) β. Note that the total pobability associated with such paths is less than o equal to β S T, since the pobability of any given path is less than o equal to β, stating with each state x in each tansition thee ae at most S possible small pobability tansitions, and thee ae T tansitions whee this can occu. Theefoe P g max g max u (p)g u (P) P u (p) β S T. We can follow the same pinciple with the MDP ˆM to conclude that (β + β) S T g max Pˆu(p)ĝ u (P). t=0 t Theefoe, we have P u (p)gu(p) Pˆ u(p)ĝ u (p) (β + 2β) S T g max (b) Fo all othe paths, we have ( )P a (x t, x t+ ) Pˆa(x t, x t+ ) ( + )P a (x t, x t+ ) whee = γ. Theefoe, β ( ) T P u (p) Pˆu(p) ( + ) T P u (p). 4
5 Moeove, g u (p) ĝ u (p) T β, then δ ( ) T [J u,t βt ] Ĵ u,t ( + ) T [J u,t + βt ] + δ 4 4 The theoem follows by consideing an appopiate choice of β. The main point (iv) says that: If exploitation is not possible, then exploation is. We show it by the following lemma. Lemma 4 Fo any x N, one of the following must hold. (a) thee exists u in M N such that Ju,T N (x) J T (x) + β, o (b) thee exists u such that the pobability that a walk of T steps will teminate in N C exceeds γ( α) g max. Poof: Let u be the policy that attains JT. If Ju N T (x) + β,t (x) J then we ae done. Suppose that Ju N,T (x) > JT (x) + β. Then we have Ju N P N (q)g N P N (p)g N,T (x) = u u (q) + u u (p) q N } }} } and Theefoe which implies q path in N path outside N J P u (q)g u (q) + P u T (x) = (q)g u (q). Ju N,T (x) Ju,T (x) = Pu N (p) g N (p) P (p)g u u } u u (p) }}} > β P N g max β() (p) > β Pu N (p). gmax α 0 g max In ode the complete the poof of Theoem fom the fou lemmas above, we have to conside the pobabilities fom two foms of failue: failue to stop the algoithm with a nea-optimal policy failue to pefom enough exploation in a timely fashion The fist point is addessed by Lemmas, 2 and 3; which establish that, if the algoithm stops, with high pobability the policy poduced is nea-optimal. The second point follows fom Lemma 4, which shows that each attempt to exploe is successful with some non negligible pobability. By applying the Chenoff bound, it can be shown that, afte a numbe of attempts that is polynomial in the quantities of inteest, exploation will occu with high pobability. 5
6 Refeences [] M. Keans and S. Singh, Nea-Optimal Reinfocement Leaning in Polynomial Time, Machine Leaning, Volume 49, Issue 2, pp , Nov
Temporal-Difference Learning
.997 Decision-Making in Lage-Scale Systems Mach 17 MIT, Sping 004 Handout #17 Lectue Note 13 1 Tempoal-Diffeence Leaning We now conside the poblem of computing an appopiate paamete, so that, given an appoximation
More informationMethod for Approximating Irrational Numbers
Method fo Appoximating Iational Numbes Eic Reichwein Depatment of Physics Univesity of Califonia, Santa Cuz June 6, 0 Abstact I will put foth an algoithm fo poducing inceasingly accuate ational appoximations
More informationLecture 18: Graph Isomorphisms
INFR11102: Computational Complexity 22/11/2018 Lectue: Heng Guo Lectue 18: Gaph Isomophisms 1 An Athu-Melin potocol fo GNI Last time we gave a simple inteactive potocol fo GNI with pivate coins. We will
More informationThe Substring Search Problem
The Substing Seach Poblem One algoithm which is used in a vaiety of applications is the family of substing seach algoithms. These algoithms allow a use to detemine if, given two chaacte stings, one is
More informationMath 301: The Erdős-Stone-Simonovitz Theorem and Extremal Numbers for Bipartite Graphs
Math 30: The Edős-Stone-Simonovitz Theoem and Extemal Numbes fo Bipatite Gaphs May Radcliffe The Edős-Stone-Simonovitz Theoem Recall, in class we poved Tuán s Gaph Theoem, namely Theoem Tuán s Theoem Let
More informationMultiple Criteria Secretary Problem: A New Approach
J. Stat. Appl. Po. 3, o., 9-38 (04 9 Jounal of Statistics Applications & Pobability An Intenational Jounal http://dx.doi.og/0.785/jsap/0303 Multiple Citeia Secetay Poblem: A ew Appoach Alaka Padhye, and
More informationStanford University CS259Q: Quantum Computing Handout 8 Luca Trevisan October 18, 2012
Stanfod Univesity CS59Q: Quantum Computing Handout 8 Luca Tevisan Octobe 8, 0 Lectue 8 In which we use the quantum Fouie tansfom to solve the peiod-finding poblem. The Peiod Finding Poblem Let f : {0,...,
More information6 Matrix Concentration Bounds
6 Matix Concentation Bounds Concentation bounds ae inequalities that bound pobabilities of deviations by a andom vaiable fom some value, often its mean. Infomally, they show the pobability that a andom
More informationProbablistically Checkable Proofs
Lectue 12 Pobablistically Checkable Poofs May 13, 2004 Lectue: Paul Beame Notes: Chis Re 12.1 Pobablisitically Checkable Poofs Oveview We know that IP = PSPACE. This means thee is an inteactive potocol
More informationLecture 28: Convergence of Random Variables and Related Theorems
EE50: Pobability Foundations fo Electical Enginees July-Novembe 205 Lectue 28: Convegence of Random Vaiables and Related Theoems Lectue:. Kishna Jagannathan Scibe: Gopal, Sudhasan, Ajay, Swamy, Kolla An
More informationSurveillance Points in High Dimensional Spaces
Société de Calcul Mathématique SA Tools fo decision help since 995 Suveillance Points in High Dimensional Spaces by Benad Beauzamy Januay 06 Abstact Let us conside any compute softwae, elying upon a lage
More information10/04/18. P [P(x)] 1 negl(n).
Mastemath, Sping 208 Into to Lattice lgs & Cypto Lectue 0 0/04/8 Lectues: D. Dadush, L. Ducas Scibe: K. de Boe Intoduction In this lectue, we will teat two main pats. Duing the fist pat we continue the
More informationA Bijective Approach to the Permutational Power of a Priority Queue
A Bijective Appoach to the Pemutational Powe of a Pioity Queue Ia M. Gessel Kuang-Yeh Wang Depatment of Mathematics Bandeis Univesity Waltham, MA 02254-9110 Abstact A pioity queue tansfoms an input pemutation
More informationGoodness-of-fit for composite hypotheses.
Section 11 Goodness-of-fit fo composite hypotheses. Example. Let us conside a Matlab example. Let us geneate 50 obsevations fom N(1, 2): X=nomnd(1,2,50,1); Then, unning a chi-squaed goodness-of-fit test
More information4/18/2005. Statistical Learning Theory
Statistical Leaning Theoy Statistical Leaning Theoy A model of supevised leaning consists of: a Envionment - Supplying a vecto x with a fixed but unknown pdf F x (x b Teache. It povides a desied esponse
More informationHypothesis Test and Confidence Interval for the Negative Binomial Distribution via Coincidence: A Case for Rare Events
Intenational Jounal of Contempoay Mathematical Sciences Vol. 12, 2017, no. 5, 243-253 HIKARI Ltd, www.m-hikai.com https://doi.og/10.12988/ijcms.2017.7728 Hypothesis Test and Confidence Inteval fo the Negative
More informationDo Managers Do Good With Other People s Money? Online Appendix
Do Manages Do Good With Othe People s Money? Online Appendix Ing-Haw Cheng Haison Hong Kelly Shue Abstact This is the Online Appendix fo Cheng, Hong and Shue 2013) containing details of the model. Datmouth
More informationFall 2014 Randomized Algorithms Oct 8, Lecture 3
Fall 204 Randomized Algoithms Oct 8, 204 Lectue 3 Pof. Fiedich Eisenband Scibes: Floian Tamè In this lectue we will be concened with linea pogamming, in paticula Clakson s Las Vegas algoithm []. The main
More informationNew problems in universal algebraic geometry illustrated by boolean equations
New poblems in univesal algebaic geomety illustated by boolean equations axiv:1611.00152v2 [math.ra] 25 Nov 2016 Atem N. Shevlyakov Novembe 28, 2016 Abstact We discuss new poblems in univesal algebaic
More informationTHE NUMBER OF TWO CONSECUTIVE SUCCESSES IN A HOPPE-PÓLYA URN
TH NUMBR OF TWO CONSCUTIV SUCCSSS IN A HOPP-PÓLYA URN LARS HOLST Depatment of Mathematics, Royal Institute of Technology S 100 44 Stocholm, Sweden -mail: lholst@math.th.se Novembe 27, 2007 Abstact In a
More informationInternet Appendix for A Bayesian Approach to Real Options: The Case of Distinguishing Between Temporary and Permanent Shocks
Intenet Appendix fo A Bayesian Appoach to Real Options: The Case of Distinguishing Between Tempoay and Pemanent Shocks Steven R. Genadie Gaduate School of Business, Stanfod Univesity Andey Malenko Gaduate
More informationFractional Zero Forcing via Three-color Forcing Games
Factional Zeo Focing via Thee-colo Focing Games Leslie Hogben Kevin F. Palmowski David E. Robeson Michael Young May 13, 2015 Abstact An -fold analogue of the positive semidefinite zeo focing pocess that
More information15 Solving the Laplace equation by Fourier method
5 Solving the Laplace equation by Fouie method I aleady intoduced two o thee dimensional heat equation, when I deived it, ecall that it taes the fom u t = α 2 u + F, (5.) whee u: [0, ) D R, D R is the
More information16 Modeling a Language by a Markov Process
K. Pommeening, Language Statistics 80 16 Modeling a Language by a Makov Pocess Fo deiving theoetical esults a common model of language is the intepetation of texts as esults of Makov pocesses. This model
More informationEM Boundary Value Problems
EM Bounday Value Poblems 10/ 9 11/ By Ilekta chistidi & Lee, Seung-Hyun A. Geneal Desciption : Maxwell Equations & Loentz Foce We want to find the equations of motion of chaged paticles. The way to do
More informationON INDEPENDENT SETS IN PURELY ATOMIC PROBABILITY SPACES WITH GEOMETRIC DISTRIBUTION. 1. Introduction. 1 r r. r k for every set E A, E \ {0},
ON INDEPENDENT SETS IN PURELY ATOMIC PROBABILITY SPACES WITH GEOMETRIC DISTRIBUTION E. J. IONASCU and A. A. STANCU Abstact. We ae inteested in constucting concete independent events in puely atomic pobability
More informationQIP Course 10: Quantum Factorization Algorithm (Part 3)
QIP Couse 10: Quantum Factoization Algoithm (Pat 3 Ryutaoh Matsumoto Nagoya Univesity, Japan Send you comments to yutaoh.matsumoto@nagoya-u.jp Septembe 2018 @ Tokyo Tech. Matsumoto (Nagoya U. QIP Couse
More informationClassical Worm algorithms (WA)
Classical Wom algoithms (WA) WA was oiginally intoduced fo quantum statistical models by Pokof ev, Svistunov and Tupitsyn (997), and late genealized to classical models by Pokof ev and Svistunov (200).
More informationApproximation Algorithms and Hardness of the k-route Cut Problem
Appoximation Algoithms and Hadness of the k-route Cut Poblem Julia Chuzhoy Yuy Makaychev Aavindan Vijayaaghavan Yuan Zhou July 10, 2011 Abstact We study the k-oute cut poblem: given an undiected edge-weighted
More informationQuantum Fourier Transform
Chapte 5 Quantum Fouie Tansfom Many poblems in physics and mathematics ae solved by tansfoming a poblem into some othe poblem with a known solution. Some notable examples ae Laplace tansfom, Legende tansfom,
More informationA Multivariate Normal Law for Turing s Formulae
A Multivaiate Nomal Law fo Tuing s Fomulae Zhiyi Zhang Depatment of Mathematics and Statistics Univesity of Noth Caolina at Chalotte Chalotte, NC 28223 Abstact This pape establishes a sufficient condition
More informationChapter 3: Theory of Modular Arithmetic 38
Chapte 3: Theoy of Modula Aithmetic 38 Section D Chinese Remainde Theoem By the end of this section you will be able to pove the Chinese Remainde Theoem apply this theoem to solve simultaneous linea conguences
More informationMath 124B February 02, 2012
Math 24B Febuay 02, 202 Vikto Gigoyan 8 Laplace s equation: popeties We have aleady encounteed Laplace s equation in the context of stationay heat conduction and wave phenomena. Recall that in two spatial
More information3.1 Random variables
3 Chapte III Random Vaiables 3 Random vaiables A sample space S may be difficult to descibe if the elements of S ae not numbes discuss how we can use a ule by which an element s of S may be associated
More information15.081J/6.251J Introduction to Mathematical Programming. Lecture 6: The Simplex Method II
15081J/6251J Intoduction to Mathematical Pogamming ectue 6: The Simplex Method II 1 Outline Revised Simplex method Slide 1 The full tableau implementation Anticycling 2 Revised Simplex Initial data: A,
More information6 PROBABILITY GENERATING FUNCTIONS
6 PROBABILITY GENERATING FUNCTIONS Cetain deivations pesented in this couse have been somewhat heavy on algeba. Fo example, detemining the expectation of the Binomial distibution (page 5.1 tuned out to
More informationCOLLAPSING WALLS THEOREM
COLLAPSING WALLS THEOREM IGOR PAK AND ROM PINCHASI Abstact. Let P R 3 be a pyamid with the base a convex polygon Q. We show that when othe faces ae collapsed (otated aound the edges onto the plane spanned
More informationMONTE CARLO STUDY OF PARTICLE TRANSPORT PROBLEM IN AIR POLLUTION. R. J. Papancheva, T. V. Gurov, I. T. Dimov
Pliska Stud. Math. Bulga. 14 (23), 17 116 STUDIA MATHEMATICA BULGARICA MOTE CARLO STUDY OF PARTICLE TRASPORT PROBLEM I AIR POLLUTIO R. J. Papancheva, T. V. Guov, I. T. Dimov Abstact. The actual tanspot
More informationRotor Blade Performance Analysis with Blade Element Momentum Theory
Available online at www.sciencediect.com ScienceDiect Enegy Pocedia 5 (7 ) 3 9 The 8 th Intenational Confeence on Applied Enegy ICAE6 Roto Blade Pefomance Analysis with Blade Element Momentum Theoy Faisal
More informationRevision of Lecture Eight
Revision of Lectue Eight Baseband equivalent system and equiements of optimal tansmit and eceive filteing: (1) achieve zeo ISI, and () maximise the eceive SNR Thee detection schemes: Theshold detection
More informationC/CS/Phys C191 Shor s order (period) finding algorithm and factoring 11/12/14 Fall 2014 Lecture 22
C/CS/Phys C9 Sho s ode (peiod) finding algoithm and factoing /2/4 Fall 204 Lectue 22 With a fast algoithm fo the uantum Fouie Tansfom in hand, it is clea that many useful applications should be possible.
More informationMATH 415, WEEK 3: Parameter-Dependence and Bifurcations
MATH 415, WEEK 3: Paamete-Dependence and Bifucations 1 A Note on Paamete Dependence We should pause to make a bief note about the ole played in the study of dynamical systems by the system s paametes.
More information1) (A B) = A B ( ) 2) A B = A. i) A A = φ i j. ii) Additional Important Properties of Sets. De Morgan s Theorems :
Additional Impotant Popeties of Sets De Mogan s Theoems : A A S S Φ, Φ S _ ( A ) A ) (A B) A B ( ) 2) A B A B Cadinality of A, A, is defined as the numbe of elements in the set A. {a,b,c} 3, { }, while
More informationBayesian Congestion Control over a Markovian Network Bandwidth Process
Bayesian Congestion Contol ove a Makovian Netwok Bandwidth Pocess Paisa Mansouifad,, Bhaska Kishnamachai, Taa Javidi Ming Hsieh Depatment of Electical Engineeing, Univesity of Southen Califonia, Los Angeles,
More informationBasic Bridge Circuits
AN7 Datafoth Copoation Page of 6 DID YOU KNOW? Samuel Hunte Chistie (784-865) was bon in London the son of James Chistie, who founded Chistie's Fine At Auctionees. Samuel studied mathematics at Tinity
More informationThe Congestion of n-cube Layout on a Rectangular Grid S.L. Bezrukov J.D. Chavez y L.H. Harper z M. Rottger U.-P. Schroeder Abstract We consider the pr
The Congestion of n-cube Layout on a Rectangula Gid S.L. Bezukov J.D. Chavez y L.H. Hape z M. Rottge U.-P. Schoede Abstact We conside the poblem of embedding the n-dimensional cube into a ectangula gid
More informationMULTILAYER PERCEPTRONS
Last updated: Nov 26, 2012 MULTILAYER PERCEPTRONS Outline 2 Combining Linea Classifies Leaning Paametes Outline 3 Combining Linea Classifies Leaning Paametes Implementing Logical Relations 4 AND and OR
More informationASTR415: Problem Set #6
ASTR45: Poblem Set #6 Cuan D. Muhlbege Univesity of Mayland (Dated: May 7, 27) Using existing implementations of the leapfog and Runge-Kutta methods fo solving coupled odinay diffeential equations, seveal
More informationInformation Retrieval Advanced IR models. Luca Bondi
Advanced IR models Luca Bondi Advanced IR models 2 (LSI) Pobabilistic Latent Semantic Analysis (plsa) Vecto Space Model 3 Stating point: Vecto Space Model Documents and queies epesented as vectos in the
More informationOn the Poisson Approximation to the Negative Hypergeometric Distribution
BULLETIN of the Malaysian Mathematical Sciences Society http://mathusmmy/bulletin Bull Malays Math Sci Soc (2) 34(2) (2011), 331 336 On the Poisson Appoximation to the Negative Hypegeometic Distibution
More informationLET a random variable x follows the two - parameter
INTERNATIONAL JOURNAL OF MATHEMATICS AND SCIENTIFIC COMPUTING ISSN: 2231-5330, VOL. 5, NO. 1, 2015 19 Shinkage Bayesian Appoach in Item - Failue Gamma Data In Pesence of Pio Point Guess Value Gyan Pakash
More informationSuggested Solutions to Homework #4 Econ 511b (Part I), Spring 2004
Suggested Solutions to Homewok #4 Econ 5b (Pat I), Sping 2004. Conside a neoclassical gowth model with valued leisue. The (epesentative) consume values steams of consumption and leisue accoding to P t=0
More informationChapter 9 Dynamic stability analysis III Lateral motion (Lectures 33 and 34)
Pof. E.G. Tulapukaa Stability and contol Chapte 9 Dynamic stability analysis Lateal motion (Lectues 33 and 34) Keywods : Lateal dynamic stability - state vaiable fom of equations, chaacteistic equation
More informationApproximation Algorithms and Hardness of the k-route Cut Problem
Appoximation Algoithms and Hadness of the k-route Cut Poblem Julia Chuzhoy Yuy Makaychev Aavindan Vijayaaghavan Yuan Zhou Decembe 14, 2011 Abstact We study the k-oute cut poblem: given an undiected edge-weighted
More informationFunctions Defined on Fuzzy Real Numbers According to Zadeh s Extension
Intenational Mathematical Foum, 3, 2008, no. 16, 763-776 Functions Defined on Fuzzy Real Numbes Accoding to Zadeh s Extension Oma A. AbuAaqob, Nabil T. Shawagfeh and Oma A. AbuGhneim 1 Mathematics Depatment,
More informationLinear Program for Partially Observable Markov Decision Processes. MS&E 339B June 9th, 2004 Erick Delage
Linea Pogam fo Patiall Obsevable Makov Decision Pocesses MS&E 339B June 9th 2004 Eick Delage Intoduction Patiall Obsevable Makov Decision Pocesses Etension of the Makov Decision Pocess to a wold with uncetaint
More informationMath 151. Rumbos Spring Solutions to Assignment #7
Math. Rumbos Sping 202 Solutions to Assignment #7. Fo each of the following, find the value of the constant c fo which the given function, p(x, is the pobability mass function (pmf of some discete andom
More informationCentral Coverage Bayes Prediction Intervals for the Generalized Pareto Distribution
Statistics Reseach Lettes Vol. Iss., Novembe Cental Coveage Bayes Pediction Intevals fo the Genealized Paeto Distibution Gyan Pakash Depatment of Community Medicine S. N. Medical College, Aga, U. P., India
More informationCompactly Supported Radial Basis Functions
Chapte 4 Compactly Suppoted Radial Basis Functions As we saw ealie, compactly suppoted functions Φ that ae tuly stictly conditionally positive definite of ode m > do not exist The compact suppot automatically
More informationApproximation Algorithms and Hardness of the k-route Cut Problem
Appoximation Algoithms and Hadness of the k-route Cut Poblem Julia Chuzhoy Yuy Makaychev Aavindan Vijayaaghavan Yuan Zhou Novembe 26, 2011 Abstact We study the k-oute cut poblem: given an undiected edge-weighted
More information2.5 The Quarter-Wave Transformer
/3/5 _5 The Quate Wave Tansfome /.5 The Quate-Wave Tansfome Reading Assignment: pp. 73-76 By now you ve noticed that a quate-wave length of tansmission line ( λ 4, β π ) appeas often in micowave engineeing
More informationUnobserved Correlation in Ascending Auctions: Example And Extensions
Unobseved Coelation in Ascending Auctions: Example And Extensions Daniel Quint Univesity of Wisconsin Novembe 2009 Intoduction In pivate-value ascending auctions, the winning bidde s willingness to pay
More informationGradient-based Neural Network for Online Solution of Lyapunov Matrix Equation with Li Activation Function
Intenational Confeence on Infomation echnology and Management Innovation (ICIMI 05) Gadient-based Neual Netwok fo Online Solution of Lyapunov Matix Equation with Li Activation unction Shiheng Wang, Shidong
More informationyou of a spring. The potential energy for a spring is given by the parabola U( x)
Small oscillations The theoy of small oscillations is an extemely impotant topic in mechanics. Conside a system that has a potential enegy diagam as below: U B C A x Thee ae thee points of stable equilibium,
More informationNotes on McCall s Model of Job Search. Timothy J. Kehoe March if job offer has been accepted. b if searching
Notes on McCall s Model of Job Seach Timothy J Kehoe Mach Fv ( ) pob( v), [, ] Choice: accept age offe o eceive b and seach again next peiod An unemployed oke solves hee max E t t y t y t if job offe has
More information763620SS STATISTICAL PHYSICS Solutions 2 Autumn 2012
763620SS STATISTICAL PHYSICS Solutions 2 Autumn 2012 1. Continuous Random Walk Conside a continuous one-dimensional andom walk. Let w(s i ds i be the pobability that the length of the i th displacement
More informationComputers and Mathematics with Applications
Computes and Mathematics with Applications 58 (009) 9 7 Contents lists available at ScienceDiect Computes and Mathematics with Applications jounal homepage: www.elsevie.com/locate/camwa Bi-citeia single
More informationand the initial value R 0 = 0, 0 = fall equivalence classes ae singletons fig; i = 1; : : : ; ng: (3) Since the tansition pobability p := P (R = j R?1
A CLASSIFICATION OF COALESCENT PROCESSES FOR HAPLOID ECHANGE- ABLE POPULATION MODELS Matin Mohle, Johannes Gutenbeg-Univesitat, Mainz and Seik Sagitov 1, Chalmes and Gotebogs Univesities, Gotebog Abstact
More informationResearch Article On Alzer and Qiu s Conjecture for Complete Elliptic Integral and Inverse Hyperbolic Tangent Function
Abstact and Applied Analysis Volume 011, Aticle ID 697547, 7 pages doi:10.1155/011/697547 Reseach Aticle On Alze and Qiu s Conjectue fo Complete Elliptic Integal and Invese Hypebolic Tangent Function Yu-Ming
More informationDirected Regression. Benjamin Van Roy Stanford University Stanford, CA Abstract
Diected Regession Yi-hao Kao Stanfod Univesity Stanfod, CA 94305 yihaoao@stanfod.edu Benjamin Van Roy Stanfod Univesity Stanfod, CA 94305 bv@stanfod.edu Xiang Yan Stanfod Univesity Stanfod, CA 94305 xyan@stanfod.edu
More informationGauss Law. Physics 231 Lecture 2-1
Gauss Law Physics 31 Lectue -1 lectic Field Lines The numbe of field lines, also known as lines of foce, ae elated to stength of the electic field Moe appopiately it is the numbe of field lines cossing
More informationQuasi-Randomness and the Distribution of Copies of a Fixed Graph
Quasi-Randomness and the Distibution of Copies of a Fixed Gaph Asaf Shapia Abstact We show that if a gaph G has the popety that all subsets of vetices of size n/4 contain the coect numbe of tiangles one
More informationOn Computing Optimal (Q, r) Replenishment Policies under Quantity Discounts
Annals of Opeations Reseach manuscipt No. will be inseted by the edito) On Computing Optimal, ) Replenishment Policies unde uantity Discounts The all - units and incemental discount cases Michael N. Katehakis
More informationFUSE Fusion Utility Sequence Estimator
FUSE Fusion Utility Sequence Estimato Belu V. Dasaathy Dynetics, Inc. P. O. Box 5500 Huntsville, AL 3584-5500 belu.d@dynetics.com Sean D. Townsend Dynetics, Inc. P. O. Box 5500 Huntsville, AL 3584-5500
More informationEnergy Levels Of Hydrogen Atom Using Ladder Operators. Ava Khamseh Supervisor: Dr. Brian Pendleton The University of Edinburgh August 2011
Enegy Levels Of Hydogen Atom Using Ladde Opeatos Ava Khamseh Supeviso: D. Bian Pendleton The Univesity of Edinbugh August 11 1 Abstact The aim of this pape is to fist use the Schödinge wavefunction methods
More information9.1 The multiplicative group of a finite field. Theorem 9.1. The multiplicative group F of a finite field is cyclic.
Chapte 9 Pimitive Roots 9.1 The multiplicative goup of a finite fld Theoem 9.1. The multiplicative goup F of a finite fld is cyclic. Remak: In paticula, if p is a pime then (Z/p) is cyclic. In fact, this
More informationMeasure Estimates of Nodal Sets of Polyharmonic Functions
Chin. Ann. Math. Se. B 39(5), 08, 97 93 DOI: 0.007/s40-08-004-6 Chinese Annals of Mathematics, Seies B c The Editoial Office of CAM and Spinge-Velag Belin Heidelbeg 08 Measue Estimates of Nodal Sets of
More informationLocalization of Eigenvalues in Small Specified Regions of Complex Plane by State Feedback Matrix
Jounal of Sciences, Islamic Republic of Ian (): - () Univesity of Tehan, ISSN - http://sciencesutaci Localization of Eigenvalues in Small Specified Regions of Complex Plane by State Feedback Matix H Ahsani
More informationLecture 2 Date:
Lectue 2 Date: 5.1.217 Definition of Some TL Paametes Examples of Tansmission Lines Tansmission Lines (contd.) Fo a lossless tansmission line the second ode diffeential equation fo phasos ae: LC 2 d I
More informationOn the integration of the equations of hydrodynamics
Uebe die Integation de hydodynamischen Gleichungen J f eine u angew Math 56 (859) -0 On the integation of the equations of hydodynamics (By A Clebsch at Calsuhe) Tanslated by D H Delphenich In a pevious
More informationHOW TO TEACH THE FUNDAMENTALS OF INFORMATION SCIENCE, CODING, DECODING AND NUMBER SYSTEMS?
6th INTERNATIONAL MULTIDISCIPLINARY CONFERENCE HOW TO TEACH THE FUNDAMENTALS OF INFORMATION SCIENCE, CODING, DECODING AND NUMBER SYSTEMS? Cecília Sitkuné Göömbei College of Nyíegyháza Hungay Abstact: The
More informationIntroduction to Mathematical Statistics Robert V. Hogg Joeseph McKean Allen T. Craig Seventh Edition
Intoduction to Mathematical Statistics Robet V. Hogg Joeseph McKean Allen T. Caig Seventh Edition Peason Education Limited Edinbugh Gate Halow Essex CM2 2JE England and Associated Companies thoughout the
More informationON THE INVERSE SIGNED TOTAL DOMINATION NUMBER IN GRAPHS. D.A. Mojdeh and B. Samadi
Opuscula Math. 37, no. 3 (017), 447 456 http://dx.doi.og/10.7494/opmath.017.37.3.447 Opuscula Mathematica ON THE INVERSE SIGNED TOTAL DOMINATION NUMBER IN GRAPHS D.A. Mojdeh and B. Samadi Communicated
More informationState tracking control for Takagi-Sugeno models
State tacing contol fo Taagi-Sugeno models Souad Bezzaoucha, Benoît Max,3,DidieMaquin,3 and José Ragot,3 Abstact This wo addesses the model efeence tacing contol poblem It aims to highlight the encouteed
More informationBayesian Analysis of Topp-Leone Distribution under Different Loss Functions and Different Priors
J. tat. Appl. Po. Lett. 3, No. 3, 9-8 (6) 9 http://dx.doi.og/.8576/jsapl/33 Bayesian Analysis of Topp-Leone Distibution unde Diffeent Loss Functions and Diffeent Pios Hummaa ultan * and. P. Ahmad Depatment
More informationINTRODUCTION. 2. Vectors in Physics 1
INTRODUCTION Vectos ae used in physics to extend the study of motion fom one dimension to two dimensions Vectos ae indispensable when a physical quantity has a diection associated with it As an example,
More informationMASSACHUSETTS INSTITUTE OF TECHNOLOGY Physics Department. Problem Set 10 Solutions. r s
MASSACHUSETTS INSTITUTE OF TECHNOLOGY Physics Depatment Physics 8.033 Decembe 5, 003 Poblem Set 10 Solutions Poblem 1 M s y x test paticle The figue above depicts the geomety of the poblem. The position
More informationANA BERRIZBEITIA, LUIS A. MEDINA, ALEXANDER C. MOLL, VICTOR H. MOLL, AND LAINE NOBLE
THE p-adic VALUATION OF STIRLING NUMBERS ANA BERRIZBEITIA, LUIS A. MEDINA, ALEXANDER C. MOLL, VICTOR H. MOLL, AND LAINE NOBLE Abstact. Let p > 2 be a pime. The p-adic valuation of Stiling numbes of the
More informationA proof of the binomial theorem
A poof of the binomial theoem If n is a natual numbe, let n! denote the poduct of the numbes,2,3,,n. So! =, 2! = 2 = 2, 3! = 2 3 = 6, 4! = 2 3 4 = 24 and so on. We also let 0! =. If n is a non-negative
More informationRelating Branching Program Size and. Formula Size over the Full Binary Basis. FB Informatik, LS II, Univ. Dortmund, Dortmund, Germany
Relating Banching Pogam Size and omula Size ove the ull Binay Basis Matin Saueho y Ingo Wegene y Ralph Wechne z y B Infomatik, LS II, Univ. Dotmund, 44 Dotmund, Gemany z ankfut, Gemany sauehof/wegene@ls.cs.uni-dotmund.de
More informationITI Introduction to Computing II
ITI 1121. Intoduction to Computing II Macel Tucotte School of Electical Engineeing and Compute Science Abstact data type: Stack Stack-based algoithms Vesion of Febuay 2, 2013 Abstact These lectue notes
More informationFailure Probability of 2-within-Consecutive-(2, 2)-out-of-(n, m): F System for Special Values of m
Jounal of Mathematics and Statistics 5 (): 0-4, 009 ISSN 549-3644 009 Science Publications Failue Pobability of -within-consecutive-(, )-out-of-(n, m): F System fo Special Values of m E.M.E.. Sayed Depatment
More information1 Notes on Order Statistics
1 Notes on Ode Statistics Fo X a andom vecto in R n with distibution F, and π S n, define X π by and F π by X π (X π(1),..., X π(n) ) F π (x 1,..., x n ) F (x π 1 (1),..., x π 1 (n)); then the distibution
More informationSecret Exponent Attacks on RSA-type Schemes with Moduli N = p r q
Secet Exponent Attacks on RSA-type Schemes with Moduli N = p q Alexande May Faculty of Compute Science, Electical Engineeing and Mathematics Univesity of Padebon 33102 Padebon, Gemany alexx@uni-padebon.de
More informationDuality between Statical and Kinematical Engineering Systems
Pape 00, Civil-Comp Ltd., Stiling, Scotland Poceedings of the Sixth Intenational Confeence on Computational Stuctues Technology, B.H.V. Topping and Z. Bittna (Editos), Civil-Comp Pess, Stiling, Scotland.
More informationDivisibility. c = bf = (ae)f = a(ef) EXAMPLE: Since 7 56 and , the Theorem above tells us that
Divisibility DEFINITION: If a and b ae integes with a 0, we say that a divides b if thee is an intege c such that b = ac. If a divides b, we also say that a is a diviso o facto of b. NOTATION: d n means
More informationCOMPUTATIONS OF ELECTROMAGNETIC FIELDS RADIATED FROM COMPLEX LIGHTNING CHANNELS
Pogess In Electomagnetics Reseach, PIER 73, 93 105, 2007 COMPUTATIONS OF ELECTROMAGNETIC FIELDS RADIATED FROM COMPLEX LIGHTNING CHANNELS T.-X. Song, Y.-H. Liu, and J.-M. Xiong School of Mechanical Engineeing
More informationRigid Body Dynamics 2. CSE169: Computer Animation Instructor: Steve Rotenberg UCSD, Winter 2018
Rigid Body Dynamics 2 CSE169: Compute Animation nstucto: Steve Rotenbeg UCSD, Winte 2018 Coss Poduct & Hat Opeato Deivative of a Rotating Vecto Let s say that vecto is otating aound the oigin, maintaining
More informationLecture 5 Solving Problems using Green s Theorem. 1. Show how Green s theorem can be used to solve general electrostatic problems 2.
Lectue 5 Solving Poblems using Geen s Theoem Today s topics. Show how Geen s theoem can be used to solve geneal electostatic poblems. Dielectics A well known application of Geen s theoem. Last time we
More informationLINEAR AND NONLINEAR ANALYSES OF A WIND-TUNNEL BALANCE
LINEAR AND NONLINEAR ANALYSES O A WIND-TUNNEL INTRODUCTION BALANCE R. Kakehabadi and R. D. Rhew NASA LaRC, Hampton, VA The NASA Langley Reseach Cente (LaRC) has been designing stain-gauge balances fo utilization
More information