Power Constrained DTNs: Risk MDP-LP Approach

Size: px
Start display at page:

Download "Power Constrained DTNs: Risk MDP-LP Approach"

Transcription

1 Power Constrined DTNs: Risk MDP-LP Approch Atul Kumr IEOR, IIT Bomby, Indi Veerrun Kvith IEOR, IIT Bomby, Indi N Hemchndr nh@iitb.c.in, IEOR, IIT Bomby, Indi. Abstrct Dely Tolernt Networks (DTNs) hve gined importnce in the recent pst, s cost-effective lterntive, in scenrios where delys cn be ccommodted. They work well in discretely connected network, where there is no direct connectivity between some/ll components of the system. But the mobility of nodes cretes occsionl contct opportunities. The rndomly moving nodes cooperte to help fixed source, in delivering messge to fr wy destintion within the given time threshold. The objective is to optimize the delivery success probbility, which turns out to be risk sensitive cost. The success probbility depends upon the contct rtes, which in turn depend upon the power used by the nodes to remin visible. The more the power used by node, the lrger is the rdius for which it is visible. However these nodes re power constrined. This leds to constrined finite horizon, Risk sensitive Mrkov Decision Process (MDP). In this pper we propose liner progrm (LP) bsed pproch to solve the corresponding dynmic progrmming equtions. This pproch enbles us in hndling the constrints. We showed using numericl simultions tht, given hrd power constrint, the solution of the constrined MDP performs significntly superior in comprison with solution obtined by optimizing joint cost. DTNs, Power lloction, Risk sensitive cost, Liner progrms, Dynmic progrmming. I. INTRODUCTION We consider lrge re with N ctive nd freely moving mobiles s in [1, wherein connectivity between different devices is vilble only occsionlly. The im is to trnsfer messge from sttic source to fixed destintion within the prescribed dedline, using the occsionl contcts between the vrious moving elements. A source trnsfers the messge to ny mobile tht comes in contct with it nd the messge is delivered to the destintion if ny one of the mobiles with messge come in contct with it. One rely mobile cnnot trnsfer the messge to nother nd this is clled the two-hop protocol (e.g., [2). And these networks re clled Dely Tolernt Networks (DTNs). Alterntively DTNs cn operte using full epidemics, i.e, the relys cn trnsfer the messge to ny other rely (see for exmple [5). The messge trnsmission performnce with full epidemics is much better; however, in terms of the power consumed it is inferior. Further there cn be flooding of messges cross the network. DTNs re operted in vrious configurtions nd using different protocols nd there is vst literture nlyzing these networks (e.g., [1, [2, [5, etc, nd references therein). One of the common techniques to nlyze these networks is using men-field dynmics (e.g., [5). This pproch is vlid in scenrios with lrge popultion. There re lso ppers tht consider the rndom system dynmics which is more ccurte with limited popultion (e.g., [1). We lso consider the rndom dynmics. An element interested in mking contct trnsmits becons (short pulses) regulrly nd contct is estblished with mobile if the lter receives one such becon. The rnge of visibility is proportionl to the power trnsmitted. Thus the more the power used, the better re contct opportunities nd the better is the probbility of successful messge trnsmission. However, the mobiles re power constrined nd the min im of this pper is to mximize the probbility of successful messge delivery, under the given power constrints. The contct process is modeled by Poisson process ([2), hence the probbility of success or equivlently probbility of delivery filure includes terms composed of powers of exponent, resulting in Risk sensitive Mrkov Decision Process (MDP) cost. Previously in DTN relted literture, such costs re hndled by exchnging the expected vlue nd the exponent using Jensen s inequlity nd the solution is obtined by optimizing bound on the objective function (e.g., [3 etc.,) Recently in [1 uthors solved the direct problem, using risk MDP pproch. However, they solved the problem using soft constrints: joint cost composed of successful delivery probbility nd the power trnsmitted is proposed, nd is optimized. While in this pper we consider the control problem with hrd constrints on the power. In [4 we showed tht the solution of risk MDP problem cn be obtined by solving corresponding Liner Progrm (LP). We obtined the solution to the power constrined DTNs by solving the LPs provided in the technicl report [4. When one is interested in hrd power control problem, the solution obtined using joint cost of [1 is obviously inferior to our direct solution of the constrined problem. More interestingly we noticed huge improvement in the performnce, becuse rndomized policy optimizes the hrd problem while the soft problem (joint cost problem) is optimized by pure policy. Thus our newly proposed LP bsed solutions re very useful in the context of hrd power constrints. II. SYSTEM MODEL AND PROBLEM DEFINITION A sttic source hs to trnsfer messge to sttic destintion within the given dedline T, nd they re sufficiently fr wy to hve ny direct communiction. The re surrounding the two hs N coopertive nd moving nodes (mobiles) tht ssist the source to deliver the messge. The source cn trnsmit only to those mobiles tht rrive in its rnge of /15/ 2015 IFIP 154

2 trnsmission. Similrly destintion cn receive informtion from only those mobiles, tht rrive within the rnge of trnsmission. And the rnge depends upon the power used for trnsmission. We sy contct occurred whenever mobile comes in the communiction rnge of the source/destintion. In lrge res with smll trnsmission rnge, the contcts re rre. In such scenrios, the contct process cn be modeled by Poisson process [2, for vriety of mobility models like rndom wlk, rndom wypoint, etc. We ssume tht the contct time is sufficient to trnsfer the messge. The source trnsfers the messge to the contcted mobile. We refer these mobiles s infected mobiles. If there is contct between the destintion nd n infected mobile within dedline T, then the messge delivery is ccomplished. Otherwise, delivery is filed. The probbility of successful delivery depends upon the power used by the source. The source derives power from bttery nd hence is power constrined. In fct, the source is provided with fixed mount of power nd it hs to ccomplish its gol utilizing the vilble power leding to hrd power constrint. The source spends power for two purposes: 1) for trnsmitting becons, to show its presence; 2) to trnsfer messge to the contcted mobiles. The power spent per trnsmitting messge could be significntly lrger thn the power spent for beconing. However beconing needs to be done t regulr instnces while the contcts re rre (in time frme of few minutes one cn t mximum mke one contct), mking the second component negligible. If the source trnsmits becons with higher power, the contct rnge increses, which further increses the contct opportunities. However the power is consumed within shorter time. On the other hnd if it trnsmits becons with lower power, it cn remin ctive for longer period, but with smller contct rnge. Thus there is n inherent trdeoff between remining ctive for longer durtion nd remining ctive with lrger contct rnge. Mobiles contcted during the erlier stges hve better chnces of delivery. Hence it might be dvntgeous to consider vrying powers for trnsmitting becons, cross the entire delivery durtion. A. Resource lloction policy We consider time slotted system, nd becons re trnsmitted with constnt power in one time slot. Without loss of generlity we ssume unit time slots. A policy represents the decisions of power levels trnsmitted in ech time slot nd the im of our pper is to obtin power policy which mximizes the probbility of successful delivery or equivlently minimizes the probbility of delivery filure. The rte of source-mobile contct Poisson process is represented by λ while tht of the destintion-mobile is given by ν. In [1, it is shown tht the contct rtes re proportionl to the power used nd hence we consider n equivlent policy in terms of (source) contct rtes. The system hs M different choices of trnsmit powers tht cn be used in ny time slot nd let Λ = {λ 0,..., λ M } represent the corresponding set of ll possible source contct rtes. Let Y t represent the contct rte chosen in time slot t. Vector Π = {π 1,, π T 1 } represents rndomized policy, where π t for ech t, is probbility distribution over Λ: π t (λ) = P rob(y t = λ) for ny λ Λ. B. Probbility of filure given policy The probbility of filure P f (Π) for given policy Π is derived in [1 nd we briefly summrize the sme here. Let X t be the number of mobiles infected t the beginning of time slot t. The sequence X t is controlled Mrkov chin, controlled by policy Π. The trnsition probbility mtrix of this controlled Mrkov chin is given by ([1): Ps λ 2 (N s 1 ), if s 1 + s 2 N p(s 1 + s 2 s 1, λ) = 0, else ( ) r with Ps λ 2 (r) := (1 e λ ) s2 e λ(r s2). s 2 Bsiclly the number infected increses by s 2 if ny s 2 mong the non-infected (N s 1 ) mobiles contct the source nd the bove is the probbility of precisely this event. A filure event occurs, when none of the X t infected mobiles contct the destintion in time slot t nd if this is true for ll the time slots. Probbility of filure is clculted by conditioning on Mrkov chin trjectory {X t } t T nd is given by (see [1 for detils): P f (Π) = E α,π [ e ν t Xt. (1) In the bove E α,π represents the expecttion under policy Π nd when the initil condition X 0 is distributed ccording to α, written s X 0 α. Here P rob(x 0 = s) = α(s). C. Totl power spent given policy The contct rte λ is proportionl to p β, where p is the trnsmitted power nd β is constnt depending upon propgtion chrcteristics of the re in which the mobiles re operting (Appendix of [1). In other words if one chooses rte λ, the power trnsmitted is proportionl to λ β. Without loss of generlity, let the constnt of proportionlity be one. Thus the totl (rndom) power spent over the T time slots equls: D. Power control problem T 1 P(Π) = Y β t. (2) t=0 The problem is to minimize the probbility of filure P f, given hrd constrint B on the verge totl power, E α,π [P, spent by the source: [ min Π Eα,Π e ν T t=0 Xt [ T 1 such tht E α,π t=0 Y β t B. (3) 155

3 Alterntively, in [1 the uthors consider joint cost depending both upon the probbility of filure P f nd e hp, term proportionl to the power [ spent: min Π Eα,Π e ν T t=0 Xt+h T 1 t=0 Y β t. (4) In the bove, h defines the weight fctor given for totl power term in the joint cost. We refer this s soft constrint (SC) problem, becuse this does not gurntee to operte within given hrd bound on the power spent. We showed in the technicl report ([4) tht the solution of constrined risk MDP problem cn be derived using the solution of n pproprite Liner Progrm (19) given in the Appendix. Thus we cn directly obtin the solution to the hrd constrint (HC) problem. The detils re given below in section III. A direct solution would obviously perform better, we compre the two solutions in section IV to determine the percentge of improvement obtined. III. LP BASED APPROACH The soft constrint (SC) problem (4) cn be cst s risk sensitive MDP problem of Appendix. The joint cost in (4), except for the logrithmic function, is similr to the risk sensitive cost J(α, Π) given s eqution (9) of Appendix, with running nd terminl costs given by rt SC (s, λ) = νs + hλ β nd rt SC (s) = νs. (5) By monotonicity, the optimiztion of cost is equivlent to optimizing the logrithm of the sme cost. The corresponding dynmic progrmming (DP) equtions of the SC problem re given by (11) fter substituting the running nd terminl costs ppropritely. One cn obtin the nlysis of the optiml policy by solving the dul LP given by (13) of Appendix. The hrd constrined (HC) problem (3) is similr to the constrined risk MDP problem (17) of Appendix. The corresponding running nd terminl costs re rt HC (s, λ) = νs nd rt HC (s) = νs, (6) while the constrint function f HC t (s, λ) = λ β for ll t. (7) Its optiml policy cn be obtined by solving Dul LP given by (19). IV. NUMERICAL ANALYSIS In [1, Lemm 2, uthors obtined structurl properties of the policy, tht optimizes the SC problem. Lemm 2 estblishes the existence of switch off threshold s off on the number infected. The optiml policy switches off (zero contct rte) the becon trnsmission, once the number infected reches the threshold. It lso showed tht the contct rte chosen below the threshold is lwys non-zero. However this sttement needs smll correction 1. For every s < s off, there exists threshold Ts (depending upon s) such tht λ t (s) λ 1 for ll t < T s nd λ t (s) = 0 for ll t T s. Thus the difference is tht, beyond s off it is lwys OFF s in [1. However below switch off popultion threshold, the ctul switch off threshold depends upon the number infected s. A. Verifiction nd comprison of SC, HC problems We begin with the verifiction of our solution to SC problem. We obtin the required solution by solving the LP (13) with the running nd terminl costs s given by (5). For simultions, we used Mtlb nd AMPL. We did most of the coding in Mtlb except for LP prt. We used AMPL to model the LP nd Gurobi solver to solve the LP. The solution x of the LP provides the optiml policy Π x s given by eqution (15) of Appendix. We then verify tht the solution stisfies the Lemm 2 of [1, fter the correction. We consider n exmple with N = 15, S = {0, 1,... 15}, T = 20, h = 20, ν = 0.1, β = 2.1 nd λ = {0, 0.1,..., 0.3}. For this exmple, the s off = s given by [1, Lemm 2. The simultion results re following the structure given by the [1, Lemm 2, s seen from the Tble I. For exmple for ll s s off, Ts = 0 nd for others it is non-zero. We hve conducted few more exmples nd verified the sme. In similr wy we obtin the solution for HC problem, now solving LP (19) with running nd terminl costs given by (6). We lso consider the constrint given by (7). We consider the following procedure for verifiction. We first solve SC problem for some vlue of weight fctor h. We compute the totl power spent by the SC problem using gin the dditionl stte component Ψ of the Appendix. Tht is, we solve SC problem lso using LP (19) with ft SC 0 for ll t. We compute the totl verge power PSC spent by the system under SC optiml policy Π SC, once gin using the eqution (18) with f t ( t ) = β t nd x = x Π SC. We then obtin the solution of HC problem with bound B set to PSC. Note here tht this procedure is only used for computing the power utilized under the lredy computed optiml policy Π SC, nd not for the purpose of constrined optimiztion. With this procedure we noticed tht both the policies consume sme power, i.e., PSC = P HC. But there is good improvement in the performnce with HC policy (see Tbles II-III). In the limited exmples tht we conducted, we observed n improvement s high s 26%. In ll these exmples we set M = 1, resulting in ON-OFF control. Thus, when the two problems obtin optiml policies with the sme power constrint, the HC solution performs superior, 1 In [1, pge 9 in the proof of Lemm2, the line fter the sentence strting with When n < n off, Q nλ β 1 < λ 1... need not be true lwys. There cn be scenrios in which f T 1 (0) < f T 1 (λ 1 ). However the lines fter tht re correct. Hence for ny n < n off, if there exists t + 1 such tht λ t+1 (n) λ 1 then for ll τ t λ τ (n) λ 1. Thus we hve the bove modifiction with Tn = t, the first t for which λ t+1 (n) λ

4 Sttes (s) Threshold time (Ts ) TABLE I VERIFICATION OF SC POLICY USING [1, LEMMA 2 T, N h ν β λ 1 P Pf S Pf H % Improvement 5, , , , TABLE II IMPROVEMENT IN P f : HC VERSUS SC, WITH EQUALLY LIKELY INITIAL CONDITIONS, α(s) = 1/( S ) FOR s S. T, N h ν β λ 1 P Pf S Pf H % Improvement 5, , , , TABLE III IMPROVEMENT IN P f : HC VERSUS SC STARTING WITH ZERO INFECTED MOBILES, α(s) = 1 {s=0} FOR s S. obviously becuse it directly solves the constrined problem. However the more interesting observtion is tht the improvement gined cn be significnt. We would now like to look t the comprison problem with different perspective. Sy we re given ny rbitrry totl verge power constrint B. Requirement is n optiml policy tht opertes within this power constrint nd which minimizes the filure probbility P f, precisely the HC problem. But if one pproches this vi SC problem, then one needs to solve the SC problem for vrious vlues of weight fctors h to obtin vrious {Pf SC (h)} h nd the corresponding totl verge powers {PSC (h)} h. Consider only those h for which totl verge power is less thn given threshold, i.e., PSC (h) B. Among these chose the best filure probbility Pf SC (h ) s the solution. Tht is, one needs to continue the serch mong SC policies, until they hit upon tht vlue of h for which the totl verge power is the mximum possible one, which is still below the given limit B. The SC solutions re known to be pure policies: π t =1 or 0, for ll t. We lso observe this to be the cse in simultions. With pure policies, the vrious choices of totl verge power would be discrete. One cn hve vrious SC solutions by considering vrious vlues of weight fctors h. However the set of ll possible totl verge powers obtined even fter exhusting the entire rnge of h, would be finite. On the other hnd HC solution is rndomized policy nd chieves the bound B with equlity, s long s it is fesible. Thus the improvement seen by directly using our LP bsed HC solution would be much more significnt thn tht demonstrted in Tbles II nd III. This effect is shown in Figure 1. This figure plots best P f performnce versus power constrint B, under both HC nd SC policies. The curve with dotted mrks, represents the best performnce fcilitted by SC solution, s function of the power constrint B, obtined by trying ll possible vlues of h. As seen from the figure, the Pf SC performnce remins constnt over rnge of power constrints B, confirming our erlier discussions. This is minly becuse the optiml policies of SC problem re lwys pure. The other curve in Figure 1 represents the P f performnce under HC policy s function of power constrint B. The entries of the Tbles II nd III correspond to the SC nd HC pir of points, where SC points re precisely the corner points of the SC curve tht re ner the HC curve. These entries lredy showed n improvement (up to 26%), nd we hve much lrger gins in the performnce t the other points (see the horizontl portions of the SC curve in Figure1). Fig. 1. P f performnce s function of power bound B B. Structurl properties We noticed from vrious exmples of the simultions tht the SC policies re ll pure policies while the HC policies re rndomized. Further, for ny given time slot t the policy suggests complete switch ON for ll sttes less thn threshold s t, rndomized switch ON-OFF t the threshold stte s = 157

5 s t nd complete switch OFF for ll sttes, s > s t. This threshold depends upon the time slot nd of course, the power constrint B. ACKNOWLEDGEMENTS: The work towrds the risk sensitive MDP problem originted with Prof. Eitn Altmn s remrk bout finding the connections between LPs nd risk sensitive MDP problems. V. CONCLUSIONS We considered power constrined Dely Tolernt Networks. We obtined optiml policies for this problem, vi the solution of n pproprite LP, fter modeling it s constrined risk sensitive MDP. The equivlence of the two is provided in the technicl report. Previously joint cost comprising of probbility of delivery filure nd term proportionl to totl power spent is considered. While in this pper we directly solve the constrined optimiztion problem. We compred the probbility of filure performnce of the DTNs under the policy so obtined, with tht of the optiml policy obtined by considering n unconstrined problem with the joint cost. We observed huge improvement. This improvement is becuse of two fctors. When the requirement is to operte optimlly within given power constrint our solution provides the optiml solution while the solution of the unconstrined problem with joint cost would be sub-optiml. Secondly nd more importntly, the optimiztion of the unconstrined risk sensitive cost results in pure policies which provide only finite choices of totl verge power spent. While our proposed constrined risk MDP solution results in optiml policies tht re rndomized nd hence provide solution, which stisfies with equlity the constrint defining the power constrint. This is true s long s the power constrint is chievble. Thus our solution performs significntly superior nd is very useful in the scenrios tht demnd for strict power constrints. APPENDIX: FINITE HORIZON RISK SENSITIVE MDP AND LINEAR PROGRAMMING Mrkov Decision Process (MDP) provides tool for solving sequentil decision mking problems in stochstic situtions. A typicl MDP consists of set S of ll possible sttes, set A of ll possible ctions nd n immedite rewrd function : S A R for ech time slot t. The terminl cost r t r T depends only upon s S. The set S nd A cn depend upon the time slot t, however we consider the sme set for ll the time indices. It is further chrcterized by trnsition function p : S A S, which defines the ction dependent stte trnsitions. Here p(s s, ) gives the probbility of the stte trnsition from s to s, when ction is chosen. We consider finite horizon problem nd let {X t } t T, {Y t } t T 1 respectively represent the trjectories of the stte nd the ction. In the lst time slot T, there is no requirement for further ction nd we only hve terminl cost. A policy = (π t, π t+1 π T 1 ) is sequence of stte dependent nd possibly rndomized ctions, given for time slots between t nd T 1. Given policy Π t nd initil condition X t = s, Π t the stte nd ction pir evolve rndomly over the time slot t < n < T, with trnsitions s governed by the following lws: q Πt (s, s, ) = P (X n = s, Y n = X n 1 = s, Y n 1 = ) = π n (s, )p(s s, ) where p(s s, ) = P (X n = s X n 1 = s, Y n 1 = ) nd π n (s, ) = P (Y n = S n = ). (8) Let E s,πt represent the expecttion opertor with initil condition X t = s nd when the policy Π t is used. Let E α,πt represent the sme expecttion opertor when the initil condition is distributed ccording to α, written s X t α. Here α(s) = P (X t = s). We re interested in optimizing the following risk sensitive objective: J t(α, Π t ) = γ 1 log (E [ α,πt e γ ) T 1 n=t rn(xn,yn)+r T (X T ). (9) The bove represents the cost to go from time slot t to T under the policy Π t with X t α. The vlue function, function of (s, t), is defined s the optiml vlue of the bove risk sensitive objective given the initil condition X t = s: V t (s) := min J t (s, Π t ) for ny s S. (10) Π t We re interested in the optiml policy Π 0 = Π (we discrd 0 in superscript when it strts from 0) tht optimizes the risk cost J 0 (s, Π 0 ), or equivlently policy tht chieves the vlue function, i.e., Π such tht V 0 (s) = J 0 (s, Π ) for ll s S. Dynmic progrmming (DP) is well known technique, tht provides solution to such control problems, nd DP equtions re given by bckwrd induction s below ([6): V T (s) = r T (s), nd for ny 0 t T 1, nd s S { [ } V t(s) = min A r t(s, ) + γ 1 log p(s s, )e γv t+1(s ). We consider the following trnsltion of the vlue function: u t (s) = e γvt(s) for ll 0 t T 1, nd s S. The DP equtions cn now be rewritten s: u t (s) = e γr T (s) nd for ny 0 t T 1, nd s S { } u t (s) = min Liner Progrmming Formultion e γrt(s,) [ p(s s, )u t+1 (s ). (11) The dynmic progrmming bsed pproch suffers from the curse of dimension. As we increse the number of sttes nd/or time epochs, the complexity of the problem increses significntly. This results in limited pplicbility of dynmic progrmming. In the context of liner MDPs, it is well known fct tht DP problem cn be reformulted s Liner Progrm (LP), under considerble generlity (see for e.g., [7, [8 in the context of infinite horizon problems). However this conversion my not solve the problem of dimension. But recent improvements in LP solvers mkes it n ttrctive lterntive. 158

6 Further nd more importntly the LP bsed pproch extends esily nd provides solutions for constrined MDPs. In technicl report ([4), we extend the LP bsed ide to finite horizon risk MDPs. In this ppendix we briefly summrize the corresponding results, while the detils nd the proofs re vilble in ([4). We hve shown in ([4) tht the solution of the unconstrined risk MDP problem (10) cn be obtined vi the solution of ny one of the two LPs, priml nd dul. In ll the discussions below, we bsorb γ into the running costs r t (.). The priml LP is given by: mx α(s)u 0 (s) (12) {u t(s)} s S,t T 1} s S subject to: u T 1(s) b s, for ll s,, u t(s) e r t(s,) p(s s, )u t+1(s ) 0 with b s, := e r T 1(s,) p(s s, )e r T (s ). for ll, s nd t T 2 In the bove {α(s); s S} is ny positive set of weights stisfying s S α(s) = 1. These cn be interpreted s the probbility distribution on initil condition. For exmple to solve (10) with t = 0, the problem with initil condition X 0 = s, one needs to set α(s) = 1 nd α(s ) = 0 for ny s s. The solution of the priml gives the trnslted vlue functions {u t (s)} while the optiml policy is directly obtined using the Dul LP: min [ e r T 1(s,) p(s s, )e r T (s ) x(t 1, s, ) (13) s S subject to: x(0, s, ) = α(s ) for ll s S x(t, s, ) = [ e rt 1(s,) p(s s, )x(t 1, s, ) s S for ll 1 t T 1 nd s S. Here gin α represents the probbility distribution on the initil condition. We hve the following results (detils in [4). We discrd the nottion superscript 0 for risk policies in the following. The bold letters represent the vectors, e.g., x = {x(t, s, )} t,s, represents fesibility vector of Dul LP (13). While s n k represents the vector s n k = [s k,, s n. Theorem 1: The following results connecting the Dul LP (13) nd the trnslted risk MDP (11) re true. 1) Fesible region nd the set of risk Policies: There is one to one correspondence between the two s below: i) For ny policy Π of risk MPD, there exists vector x Π which stisfies ll the constrints of Dul LP (13). The fesible vector is given by the eqution (see (8)): x Π(0, s 0, 0) = α(s 0)π 0(s 0, 0) for ll s 0 S, 0 A, x Π(t, s t, t) = α(s t 1 0)e n=0 rn(sn,n) Π t n=0 q Π (s n, n s n 1, n 1) t 1 0,s t 1 0 for ll s t S, t A, nd 1 t < T. (14) ii) Given vector x in the fesibility region of Dul LP, define policy Π x using the following rule: π x,t (s, ) := x(t, s, ) x(t, s, ) for ll s S, nd A. (15) The vector x Πx defined by eqution (14) of point (i) is gin in the fesibility region nd equls x. 2) Optiml policies nd solutions: () If x is n optiml solution of the Dul LP, then Π x defined by (15) is n optiml policy for risk MDP. 3) Expecttion t optiml Policy: For ny fesible point x of Dul LP nd for ny integrble function f, s t, t x(t, s t, t )f(s t, t ) Constrined risk MDP = E Πx [ e t 1 n=0 rn(xn,yn) f(x t, Y t ). (16) We now consider constrined MDP problem (detils re in [4), with n dditionl constrint s given below: min Π J 0(α, Π) (17) Subject to: t Eα,Π [f t (X t, Y t ) B, for some set of integrble function {f t }, initil distribution α nd bound B. The eqution (16) of Theorem 1 could hve been useful in obtining the expecttion defining the constrint, but for the extr fctor Ψ 1 t with Ψ t := e t 1 n=0 rn(xn,yn), s seen from the right hnd side of the eqution (16). We propose to dd Ψ t s dditionl stte component to the originl Mrkov process {X t } to tckle this problem. We consider two component stte evolution {(X t, Ψ t )} nd the corresponding probbility trnsition mtrix depends explicitly upon time index s below: p t+1 (s, ψ t+1 s, ψ t, ) = 1 {ψ t+1 =ψ te r t (s,) }p(s s, ). With the introduction of the new stte component, for ny Dul LP fesible point x we hve: s t,ψ t, t x(t, s t, ψ t, t) ψ t f(s t, t) = E Πx [f(x t, Y t). (18) 159

7 Thus one cn obtin optiml policy of constrined risk MDP (17) by considering n dditionl stte component nd by dding n extr constrint to the Dul LP (13) s below: min [ e r T 1(s,) p(s s, )e r T (s ) x(t 1, s, ) (19) s subject to: x(t, s, ) = ψ t x(t, s, ψ t, ) x(0, s, ψ 0, ) = α(s)1 {ψ0 =1} for ll s, ψ 0 x(t, s, ψ t, ) = t s,ψ t,,s,ψ t 1 e rt 1(s,) p(s, ψ t s, ψ t 1, )x(t 1, s, ψ t 1, ) x(t, s, ψ t, )ψ tf t(s, ) B. for ll 1 t T 1 nd s, ψ t nd We would like to mention here tht ψ 0 is lwys initilized to one, i.e., ψ 0 = 1, ψ 1 cn tke t mximum S A vlues while ψ t for ny t cn tke t mximum S t A t possible vlues. There will lso be considerble deletions if the concerned mpping ( t 0, s t 0) e t n=0 rn(sn,n) is not one-one. One needs to consider this time dependent stte spce while solving the Dul LP given bove nd we omit the discussion of these obvious detils. REFERENCES [1 E. Altmn, V. Kvith, F. De Pellegrini, V. Kmble, nd V. Borkr, Risk sensitive optiml control frmework pplied to dely tolernt networks, in INFOCOM, 2011 Proceedings IEEE. IEEE, 2011, pp [2 R. Groenevelt, P. Nin, nd G. Koole, Messge dely in mnet, in ACM SIGMETRICS Performnce Evlution Review, vol. 33, no. 1. ACM, 2005, pp [3 E. Altmn, T. Bşr, nd F. De Pellegrini, Optiml monotone forwrding policies in dely tolernt mobile d-hoc networks, Performnce Evlution, vol. 67, no. 4, pp , [4 Atul Kumr, Veerrun Kvith nd N. Hemchndr, Finite horizon risk sensitive MDP nd liner progrmming, Mnuscript under preprtion. Technicl report vilble t [5 Khouzni, M. H. R., Eshghi, S., Srkr, S., Shroff, N. B., nd Venktesh, S. S. (2012, June). Optiml energy-wre epidemic routing in DTNs, In Proceedings of the thirteenth ACM interntionl symposium on Mobile Ad Hoc Networking nd Computing (pp ). ACM. [6 S. P. Corluppi nd S. I. Mrcus, Risk-sensitive queueing, in Proceedings of the Annul Allerton Conference on Communiction Control nd Computing, vol. 35. Citeseer, 1997, pp [7 M. L. Putermn, Mrkov decision processes: discrete stochstic dynmic progrmming. John Wiley & Sons, [8 Eitn Altmn, Constrined Mrkov decision processes, volume 7, 1999, CRC Press. 160

Finite Horizon Risk Sensitive MDP and Linear Programming

Finite Horizon Risk Sensitive MDP and Linear Programming Finite Horizon Risk Sensitive MDP nd Liner Progrmming Atul Kumr, Veerrun Kvith nd N. Hemchndr IEOR, Indin Institute of Technology Bomby, Indi Abstrct In the context of stndrd Mrkov decision processes (MDPs),

More information

Reinforcement Learning

Reinforcement Learning Reinforcement Lerning Tom Mitchell, Mchine Lerning, chpter 13 Outline Introduction Comprison with inductive lerning Mrkov Decision Processes: the model Optiml policy: The tsk Q Lerning: Q function Algorithm

More information

Duality # Second iteration for HW problem. Recall our LP example problem we have been working on, in equality form, is given below.

Duality # Second iteration for HW problem. Recall our LP example problem we have been working on, in equality form, is given below. Dulity #. Second itertion for HW problem Recll our LP emple problem we hve been working on, in equlity form, is given below.,,,, 8 m F which, when written in slightly different form, is 8 F Recll tht we

More information

Reinforcement learning II

Reinforcement learning II CS 1675 Introduction to Mchine Lerning Lecture 26 Reinforcement lerning II Milos Huskrecht milos@cs.pitt.edu 5329 Sennott Squre Reinforcement lerning Bsics: Input x Lerner Output Reinforcement r Critic

More information

Administrivia CSE 190: Reinforcement Learning: An Introduction

Administrivia CSE 190: Reinforcement Learning: An Introduction Administrivi CSE 190: Reinforcement Lerning: An Introduction Any emil sent to me bout the course should hve CSE 190 in the subject line! Chpter 4: Dynmic Progrmming Acknowledgment: A good number of these

More information

1 Online Learning and Regret Minimization

1 Online Learning and Regret Minimization 2.997 Decision-Mking in Lrge-Scle Systems My 10 MIT, Spring 2004 Hndout #29 Lecture Note 24 1 Online Lerning nd Regret Minimiztion In this lecture, we consider the problem of sequentil decision mking in

More information

Tests for the Ratio of Two Poisson Rates

Tests for the Ratio of Two Poisson Rates Chpter 437 Tests for the Rtio of Two Poisson Rtes Introduction The Poisson probbility lw gives the probbility distribution of the number of events occurring in specified intervl of time or spce. The Poisson

More information

Properties of Integrals, Indefinite Integrals. Goals: Definition of the Definite Integral Integral Calculations using Antiderivatives

Properties of Integrals, Indefinite Integrals. Goals: Definition of the Definite Integral Integral Calculations using Antiderivatives Block #6: Properties of Integrls, Indefinite Integrls Gols: Definition of the Definite Integrl Integrl Clcultions using Antiderivtives Properties of Integrls The Indefinite Integrl 1 Riemnn Sums - 1 Riemnn

More information

Online Supplements to Performance-Based Contracts for Outpatient Medical Services

Online Supplements to Performance-Based Contracts for Outpatient Medical Services Jing, Png nd Svin: Performnce-bsed Contrcts Article submitted to Mnufcturing & Service Opertions Mngement; mnuscript no. MSOM-11-270.R2 1 Online Supplements to Performnce-Bsed Contrcts for Outptient Medicl

More information

19 Optimal behavior: Game theory

19 Optimal behavior: Game theory Intro. to Artificil Intelligence: Dle Schuurmns, Relu Ptrscu 1 19 Optiml behvior: Gme theory Adversril stte dynmics hve to ccount for worst cse Compute policy π : S A tht mximizes minimum rewrd Let S (,

More information

Improper Integrals. Type I Improper Integrals How do we evaluate an integral such as

Improper Integrals. Type I Improper Integrals How do we evaluate an integral such as Improper Integrls Two different types of integrls cn qulify s improper. The first type of improper integrl (which we will refer to s Type I) involves evluting n integrl over n infinite region. In the grph

More information

Acceptance Sampling by Attributes

Acceptance Sampling by Attributes Introduction Acceptnce Smpling by Attributes Acceptnce smpling is concerned with inspection nd decision mking regrding products. Three spects of smpling re importnt: o Involves rndom smpling of n entire

More information

Bellman Optimality Equation for V*

Bellman Optimality Equation for V* Bellmn Optimlity Eqution for V* The vlue of stte under n optiml policy must equl the expected return for the best ction from tht stte: V (s) mx Q (s,) A(s) mx A(s) mx A(s) Er t 1 V (s t 1 ) s t s, t s

More information

Review of Calculus, cont d

Review of Calculus, cont d Jim Lmbers MAT 460 Fll Semester 2009-10 Lecture 3 Notes These notes correspond to Section 1.1 in the text. Review of Clculus, cont d Riemnn Sums nd the Definite Integrl There re mny cses in which some

More information

7.2 The Definite Integral

7.2 The Definite Integral 7.2 The Definite Integrl the definite integrl In the previous section, it ws found tht if function f is continuous nd nonnegtive, then the re under the grph of f on [, b] is given by F (b) F (), where

More information

The Regulated and Riemann Integrals

The Regulated and Riemann Integrals Chpter 1 The Regulted nd Riemnn Integrls 1.1 Introduction We will consider severl different pproches to defining the definite integrl f(x) dx of function f(x). These definitions will ll ssign the sme vlue

More information

2D1431 Machine Learning Lab 3: Reinforcement Learning

2D1431 Machine Learning Lab 3: Reinforcement Learning 2D1431 Mchine Lerning Lb 3: Reinforcement Lerning Frnk Hoffmnn modified by Örjn Ekeberg December 7, 2004 1 Introduction In this lb you will lern bout dynmic progrmming nd reinforcement lerning. It is ssumed

More information

Lecture 1. Functional series. Pointwise and uniform convergence.

Lecture 1. Functional series. Pointwise and uniform convergence. 1 Introduction. Lecture 1. Functionl series. Pointwise nd uniform convergence. In this course we study mongst other things Fourier series. The Fourier series for periodic function f(x) with period 2π is

More information

Continuous Random Variables

Continuous Random Variables STAT/MATH 395 A - PROBABILITY II UW Winter Qurter 217 Néhémy Lim Continuous Rndom Vribles Nottion. The indictor function of set S is rel-vlued function defined by : { 1 if x S 1 S (x) if x S Suppose tht

More information

Generalized Fano and non-fano networks

Generalized Fano and non-fano networks Generlized Fno nd non-fno networks Nildri Ds nd Brijesh Kumr Ri Deprtment of Electronics nd Electricl Engineering Indin Institute of Technology Guwhti, Guwhti, Assm, Indi Emil: {d.nildri, bkri}@iitg.ernet.in

More information

UNIT 1 FUNCTIONS AND THEIR INVERSES Lesson 1.4: Logarithmic Functions as Inverses Instruction

UNIT 1 FUNCTIONS AND THEIR INVERSES Lesson 1.4: Logarithmic Functions as Inverses Instruction Lesson : Logrithmic Functions s Inverses Prerequisite Skills This lesson requires the use of the following skills: determining the dependent nd independent vribles in n exponentil function bsed on dt from

More information

CBE 291b - Computation And Optimization For Engineers

CBE 291b - Computation And Optimization For Engineers The University of Western Ontrio Fculty of Engineering Science Deprtment of Chemicl nd Biochemicl Engineering CBE 9b - Computtion And Optimiztion For Engineers Mtlb Project Introduction Prof. A. Jutn Jn

More information

Driving Cycle Construction of City Road for Hybrid Bus Based on Markov Process Deng Pan1, a, Fengchun Sun1,b*, Hongwen He1, c, Jiankun Peng1, d

Driving Cycle Construction of City Road for Hybrid Bus Based on Markov Process Deng Pan1, a, Fengchun Sun1,b*, Hongwen He1, c, Jiankun Peng1, d Interntionl Industril Informtics nd Computer Engineering Conference (IIICEC 15) Driving Cycle Construction of City Rod for Hybrid Bus Bsed on Mrkov Process Deng Pn1,, Fengchun Sun1,b*, Hongwen He1, c,

More information

SUMMER KNOWHOW STUDY AND LEARNING CENTRE

SUMMER KNOWHOW STUDY AND LEARNING CENTRE SUMMER KNOWHOW STUDY AND LEARNING CENTRE Indices & Logrithms 2 Contents Indices.2 Frctionl Indices.4 Logrithms 6 Exponentil equtions. Simplifying Surds 13 Opertions on Surds..16 Scientific Nottion..18

More information

f(x) dx, If one of these two conditions is not met, we call the integral improper. Our usual definition for the value for the definite integral

f(x) dx, If one of these two conditions is not met, we call the integral improper. Our usual definition for the value for the definite integral Improper Integrls Every time tht we hve evluted definite integrl such s f(x) dx, we hve mde two implicit ssumptions bout the integrl:. The intervl [, b] is finite, nd. f(x) is continuous on [, b]. If one

More information

Advanced Calculus: MATH 410 Notes on Integrals and Integrability Professor David Levermore 17 October 2004

Advanced Calculus: MATH 410 Notes on Integrals and Integrability Professor David Levermore 17 October 2004 Advnced Clculus: MATH 410 Notes on Integrls nd Integrbility Professor Dvid Levermore 17 October 2004 1. Definite Integrls In this section we revisit the definite integrl tht you were introduced to when

More information

Lecture 14: Quadrature

Lecture 14: Quadrature Lecture 14: Qudrture This lecture is concerned with the evlution of integrls fx)dx 1) over finite intervl [, b] The integrnd fx) is ssumed to be rel-vlues nd smooth The pproximtion of n integrl by numericl

More information

Math& 152 Section Integration by Parts

Math& 152 Section Integration by Parts Mth& 5 Section 7. - Integrtion by Prts Integrtion by prts is rule tht trnsforms the integrl of the product of two functions into other (idelly simpler) integrls. Recll from Clculus I tht given two differentible

More information

W. We shall do so one by one, starting with I 1, and we shall do it greedily, trying

W. We shall do so one by one, starting with I 1, and we shall do it greedily, trying Vitli covers 1 Definition. A Vitli cover of set E R is set V of closed intervls with positive length so tht, for every δ > 0 nd every x E, there is some I V with λ(i ) < δ nd x I. 2 Lemm (Vitli covering)

More information

( dg. ) 2 dt. + dt. dt j + dh. + dt. r(t) dt. Comparing this equation with the one listed above for the length of see that

( dg. ) 2 dt. + dt. dt j + dh. + dt. r(t) dt. Comparing this equation with the one listed above for the length of see that Arc Length of Curves in Three Dimensionl Spce If the vector function r(t) f(t) i + g(t) j + h(t) k trces out the curve C s t vries, we cn mesure distnces long C using formul nerly identicl to one tht we

More information

SOLUTIONS FOR ADMISSIONS TEST IN MATHEMATICS, COMPUTER SCIENCE AND JOINT SCHOOLS WEDNESDAY 5 NOVEMBER 2014

SOLUTIONS FOR ADMISSIONS TEST IN MATHEMATICS, COMPUTER SCIENCE AND JOINT SCHOOLS WEDNESDAY 5 NOVEMBER 2014 SOLUTIONS FOR ADMISSIONS TEST IN MATHEMATICS, COMPUTER SCIENCE AND JOINT SCHOOLS WEDNESDAY 5 NOVEMBER 014 Mrk Scheme: Ech prt of Question 1 is worth four mrks which re wrded solely for the correct nswer.

More information

MATH 144: Business Calculus Final Review

MATH 144: Business Calculus Final Review MATH 144: Business Clculus Finl Review 1 Skills 1. Clculte severl limits. 2. Find verticl nd horizontl symptotes for given rtionl function. 3. Clculte derivtive by definition. 4. Clculte severl derivtives

More information

Exam 2, Mathematics 4701, Section ETY6 6:05 pm 7:40 pm, March 31, 2016, IH-1105 Instructor: Attila Máté 1

Exam 2, Mathematics 4701, Section ETY6 6:05 pm 7:40 pm, March 31, 2016, IH-1105 Instructor: Attila Máté 1 Exm, Mthemtics 471, Section ETY6 6:5 pm 7:4 pm, Mrch 1, 16, IH-115 Instructor: Attil Máté 1 17 copies 1. ) Stte the usul sufficient condition for the fixed-point itertion to converge when solving the eqution

More information

Lecture 3 ( ) (translated and slightly adapted from lecture notes by Martin Klazar)

Lecture 3 ( ) (translated and slightly adapted from lecture notes by Martin Klazar) Lecture 3 (5.3.2018) (trnslted nd slightly dpted from lecture notes by Mrtin Klzr) Riemnn integrl Now we define precisely the concept of the re, in prticulr, the re of figure U(, b, f) under the grph of

More information

Math 1B, lecture 4: Error bounds for numerical methods

Math 1B, lecture 4: Error bounds for numerical methods Mth B, lecture 4: Error bounds for numericl methods Nthn Pflueger 4 September 0 Introduction The five numericl methods descried in the previous lecture ll operte by the sme principle: they pproximte the

More information

APPROXIMATE INTEGRATION

APPROXIMATE INTEGRATION APPROXIMATE INTEGRATION. Introduction We hve seen tht there re functions whose nti-derivtives cnnot be expressed in closed form. For these resons ny definite integrl involving these integrnds cnnot be

More information

ECO 317 Economics of Uncertainty Fall Term 2007 Notes for lectures 4. Stochastic Dominance

ECO 317 Economics of Uncertainty Fall Term 2007 Notes for lectures 4. Stochastic Dominance Generl structure ECO 37 Economics of Uncertinty Fll Term 007 Notes for lectures 4. Stochstic Dominnce Here we suppose tht the consequences re welth mounts denoted by W, which cn tke on ny vlue between

More information

Time Optimal Control of the Brockett Integrator

Time Optimal Control of the Brockett Integrator Milno (Itly) August 8 - September, 011 Time Optiml Control of the Brockett Integrtor S. Sinh Deprtment of Mthemtics, IIT Bomby, Mumbi, Indi (emil : sunnysphs4891@gmil.com) Abstrct: The Brockett integrtor

More information

Numerical Analysis: Trapezoidal and Simpson s Rule

Numerical Analysis: Trapezoidal and Simpson s Rule nd Simpson s Mthemticl question we re interested in numericlly nswering How to we evlute I = f (x) dx? Clculus tells us tht if F(x) is the ntiderivtive of function f (x) on the intervl [, b], then I =

More information

An approximation to the arithmetic-geometric mean. G.J.O. Jameson, Math. Gazette 98 (2014), 85 95

An approximation to the arithmetic-geometric mean. G.J.O. Jameson, Math. Gazette 98 (2014), 85 95 An pproximtion to the rithmetic-geometric men G.J.O. Jmeson, Mth. Gzette 98 (4), 85 95 Given positive numbers > b, consider the itertion given by =, b = b nd n+ = ( n + b n ), b n+ = ( n b n ) /. At ech

More information

Decision Networks. CS 188: Artificial Intelligence Fall Example: Decision Networks. Decision Networks. Decisions as Outcome Trees

Decision Networks. CS 188: Artificial Intelligence Fall Example: Decision Networks. Decision Networks. Decisions as Outcome Trees CS 188: Artificil Intelligence Fll 2011 Decision Networks ME: choose the ction which mximizes the expected utility given the evidence mbrell Lecture 17: Decision Digrms 10/27/2011 Cn directly opertionlize

More information

THE EXISTENCE-UNIQUENESS THEOREM FOR FIRST-ORDER DIFFERENTIAL EQUATIONS.

THE EXISTENCE-UNIQUENESS THEOREM FOR FIRST-ORDER DIFFERENTIAL EQUATIONS. THE EXISTENCE-UNIQUENESS THEOREM FOR FIRST-ORDER DIFFERENTIAL EQUATIONS RADON ROSBOROUGH https://intuitiveexplntionscom/picrd-lindelof-theorem/ This document is proof of the existence-uniqueness theorem

More information

Cf. Linn Sennott, Stochastic Dynamic Programming and the Control of Queueing Systems, Wiley Series in Probability & Statistics, 1999.

Cf. Linn Sennott, Stochastic Dynamic Programming and the Control of Queueing Systems, Wiley Series in Probability & Statistics, 1999. Cf. Linn Sennott, Stochstic Dynmic Progrmming nd the Control of Queueing Systems, Wiley Series in Probbility & Sttistics, 1999. D.L.Bricker, 2001 Dept of Industril Engineering The University of Iow MDP

More information

CS 188 Introduction to Artificial Intelligence Fall 2018 Note 7

CS 188 Introduction to Artificial Intelligence Fall 2018 Note 7 CS 188 Introduction to Artificil Intelligence Fll 2018 Note 7 These lecture notes re hevily bsed on notes originlly written by Nikhil Shrm. Decision Networks In the third note, we lerned bout gme trees

More information

Chapter 4 Contravariance, Covariance, and Spacetime Diagrams

Chapter 4 Contravariance, Covariance, and Spacetime Diagrams Chpter 4 Contrvrince, Covrince, nd Spcetime Digrms 4. The Components of Vector in Skewed Coordintes We hve seen in Chpter 3; figure 3.9, tht in order to show inertil motion tht is consistent with the Lorentz

More information

Estimation of Binomial Distribution in the Light of Future Data

Estimation of Binomial Distribution in the Light of Future Data British Journl of Mthemtics & Computer Science 102: 1-7, 2015, Article no.bjmcs.19191 ISSN: 2231-0851 SCIENCEDOMAIN interntionl www.sciencedomin.org Estimtion of Binomil Distribution in the Light of Future

More information

DIRECT CURRENT CIRCUITS

DIRECT CURRENT CIRCUITS DRECT CURRENT CUTS ELECTRC POWER Consider the circuit shown in the Figure where bttery is connected to resistor R. A positive chrge dq will gin potentil energy s it moves from point to point b through

More information

Goals: Determine how to calculate the area described by a function. Define the definite integral. Explore the relationship between the definite

Goals: Determine how to calculate the area described by a function. Define the definite integral. Explore the relationship between the definite Unit #8 : The Integrl Gols: Determine how to clculte the re described by function. Define the definite integrl. Eplore the reltionship between the definite integrl nd re. Eplore wys to estimte the definite

More information

Chapters 4 & 5 Integrals & Applications

Chapters 4 & 5 Integrals & Applications Contents Chpters 4 & 5 Integrls & Applictions Motivtion to Chpters 4 & 5 2 Chpter 4 3 Ares nd Distnces 3. VIDEO - Ares Under Functions............................................ 3.2 VIDEO - Applictions

More information

5.7 Improper Integrals

5.7 Improper Integrals 458 pplictions of definite integrls 5.7 Improper Integrls In Section 5.4, we computed the work required to lift pylod of mss m from the surfce of moon of mss nd rdius R to height H bove the surfce of the

More information

MAA 4212 Improper Integrals

MAA 4212 Improper Integrals Notes by Dvid Groisser, Copyright c 1995; revised 2002, 2009, 2014 MAA 4212 Improper Integrls The Riemnn integrl, while perfectly well-defined, is too restrictive for mny purposes; there re functions which

More information

Monte Carlo method in solving numerical integration and differential equation

Monte Carlo method in solving numerical integration and differential equation Monte Crlo method in solving numericl integrtion nd differentil eqution Ye Jin Chemistry Deprtment Duke University yj66@duke.edu Abstrct: Monte Crlo method is commonly used in rel physics problem. The

More information

Calculus I-II Review Sheet

Calculus I-II Review Sheet Clculus I-II Review Sheet 1 Definitions 1.1 Functions A function is f is incresing on n intervl if x y implies f(x) f(y), nd decresing if x y implies f(x) f(y). It is clled monotonic if it is either incresing

More information

How to simulate Turing machines by invertible one-dimensional cellular automata

How to simulate Turing machines by invertible one-dimensional cellular automata How to simulte Turing mchines by invertible one-dimensionl cellulr utomt Jen-Christophe Dubcq Déprtement de Mthémtiques et d Informtique, École Normle Supérieure de Lyon, 46, llée d Itlie, 69364 Lyon Cedex

More information

Chapter 5 : Continuous Random Variables

Chapter 5 : Continuous Random Variables STAT/MATH 395 A - PROBABILITY II UW Winter Qurter 216 Néhémy Lim Chpter 5 : Continuous Rndom Vribles Nottions. N {, 1, 2,...}, set of nturl numbers (i.e. ll nonnegtive integers); N {1, 2,...}, set of ll

More information

I1 = I2 I1 = I2 + I3 I1 + I2 = I3 + I4 I 3

I1 = I2 I1 = I2 + I3 I1 + I2 = I3 + I4 I 3 2 The Prllel Circuit Electric Circuits: Figure 2- elow show ttery nd multiple resistors rrnged in prllel. Ech resistor receives portion of the current from the ttery sed on its resistnce. The split is

More information

3.1 Exponential Functions and Their Graphs

3.1 Exponential Functions and Their Graphs . Eponentil Functions nd Their Grphs Sllbus Objective: 9. The student will sketch the grph of eponentil, logistic, or logrithmic function. 9. The student will evlute eponentil or logrithmic epressions.

More information

ARITHMETIC OPERATIONS. The real numbers have the following properties: a b c ab ac

ARITHMETIC OPERATIONS. The real numbers have the following properties: a b c ab ac REVIEW OF ALGEBRA Here we review the bsic rules nd procedures of lgebr tht you need to know in order to be successful in clculus. ARITHMETIC OPERATIONS The rel numbers hve the following properties: b b

More information

Improper Integrals, and Differential Equations

Improper Integrals, and Differential Equations Improper Integrls, nd Differentil Equtions October 22, 204 5.3 Improper Integrls Previously, we discussed how integrls correspond to res. More specificlly, we sid tht for function f(x), the region creted

More information

{ } = E! & $ " k r t +k +1

{ } = E! & $  k r t +k +1 Chpter 4: Dynmic Progrmming Objectives of this chpter: Overview of collection of clssicl solution methods for MDPs known s dynmic progrmming (DP) Show how DP cn be used to compute vlue functions, nd hence,

More information

Vyacheslav Telnin. Search for New Numbers.

Vyacheslav Telnin. Search for New Numbers. Vycheslv Telnin Serch for New Numbers. 1 CHAPTER I 2 I.1 Introduction. In 1984, in the first issue for tht yer of the Science nd Life mgzine, I red the rticle "Non-Stndrd Anlysis" by V. Uspensky, in which

More information

Chapter 4: Dynamic Programming

Chapter 4: Dynamic Programming Chpter 4: Dynmic Progrmming Objectives of this chpter: Overview of collection of clssicl solution methods for MDPs known s dynmic progrmming (DP) Show how DP cn be used to compute vlue functions, nd hence,

More information

Emission of K -, L - and M - Auger Electrons from Cu Atoms. Abstract

Emission of K -, L - and M - Auger Electrons from Cu Atoms. Abstract Emission of K -, L - nd M - uger Electrons from Cu toms Mohmed ssd bdel-rouf Physics Deprtment, Science College, UEU, l in 17551, United rb Emirtes ssd@ueu.c.e bstrct The emission of uger electrons from

More information

State space systems analysis (continued) Stability. A. Definitions A system is said to be Asymptotically Stable (AS) when it satisfies

State space systems analysis (continued) Stability. A. Definitions A system is said to be Asymptotically Stable (AS) when it satisfies Stte spce systems nlysis (continued) Stbility A. Definitions A system is sid to be Asymptoticlly Stble (AS) when it stisfies ut () = 0, t > 0 lim xt () 0. t A system is AS if nd only if the impulse response

More information

1 Probability Density Functions

1 Probability Density Functions Lis Yn CS 9 Continuous Distributions Lecture Notes #9 July 6, 28 Bsed on chpter by Chris Piech So fr, ll rndom vribles we hve seen hve been discrete. In ll the cses we hve seen in CS 9, this ment tht our

More information

Chapter 0. What is the Lebesgue integral about?

Chapter 0. What is the Lebesgue integral about? Chpter 0. Wht is the Lebesgue integrl bout? The pln is to hve tutoril sheet ech week, most often on Fridy, (to be done during the clss) where you will try to get used to the ides introduced in the previous

More information

CS 188: Artificial Intelligence Spring 2007

CS 188: Artificial Intelligence Spring 2007 CS 188: Artificil Intelligence Spring 2007 Lecture 3: Queue-Bsed Serch 1/23/2007 Srini Nrynn UC Berkeley Mny slides over the course dpted from Dn Klein, Sturt Russell or Andrew Moore Announcements Assignment

More information

Using QM for Windows. Using QM for Windows. Using QM for Windows LEARNING OBJECTIVES. Solving Flair Furniture s LP Problem

Using QM for Windows. Using QM for Windows. Using QM for Windows LEARNING OBJECTIVES. Solving Flair Furniture s LP Problem LEARNING OBJECTIVES Vlu%on nd pricing (November 5, 2013) Lecture 11 Liner Progrmming (prt 2) 10/8/16, 2:46 AM Olivier J. de Jong, LL.M., MM., MBA, CFD, CFFA, AA www.olivierdejong.com Solving Flir Furniture

More information

New data structures to reduce data size and search time

New data structures to reduce data size and search time New dt structures to reduce dt size nd serch time Tsuneo Kuwbr Deprtment of Informtion Sciences, Fculty of Science, Kngw University, Hirtsuk-shi, Jpn FIT2018 1D-1, No2, pp1-4 Copyright (c)2018 by The Institute

More information

Stuff You Need to Know From Calculus

Stuff You Need to Know From Calculus Stuff You Need to Know From Clculus For the first time in the semester, the stuff we re doing is finlly going to look like clculus (with vector slnt, of course). This mens tht in order to succeed, you

More information

221B Lecture Notes WKB Method

221B Lecture Notes WKB Method Clssicl Limit B Lecture Notes WKB Method Hmilton Jcobi Eqution We strt from the Schrödinger eqution for single prticle in potentil i h t ψ x, t = [ ] h m + V x ψ x, t. We cn rewrite this eqution by using

More information

Name Solutions to Test 3 November 8, 2017

Name Solutions to Test 3 November 8, 2017 Nme Solutions to Test 3 November 8, 07 This test consists of three prts. Plese note tht in prts II nd III, you cn skip one question of those offered. Some possibly useful formuls cn be found below. Brrier

More information

For the percentage of full time students at RCC the symbols would be:

For the percentage of full time students at RCC the symbols would be: Mth 17/171 Chpter 7- ypothesis Testing with One Smple This chpter is s simple s the previous one, except it is more interesting In this chpter we will test clims concerning the sme prmeters tht we worked

More information

Artificial Intelligence Markov Decision Problems

Artificial Intelligence Markov Decision Problems rtificil Intelligence Mrkov eciion Problem ilon - briefly mentioned in hpter Ruell nd orvig - hpter 7 Mrkov eciion Problem; pge of Mrkov eciion Problem; pge of exmple: probbilitic blockworld ction outcome

More information

Taylor Polynomial Inequalities

Taylor Polynomial Inequalities Tylor Polynomil Inequlities Ben Glin September 17, 24 Abstrct There re instnces where we my wish to pproximte the vlue of complicted function round given point by constructing simpler function such s polynomil

More information

1.9 C 2 inner variations

1.9 C 2 inner variations 46 CHAPTER 1. INDIRECT METHODS 1.9 C 2 inner vritions So fr, we hve restricted ttention to liner vritions. These re vritions of the form vx; ǫ = ux + ǫφx where φ is in some liner perturbtion clss P, for

More information

Module 6: LINEAR TRANSFORMATIONS

Module 6: LINEAR TRANSFORMATIONS Module 6: LINEAR TRANSFORMATIONS. Trnsformtions nd mtrices Trnsformtions re generliztions of functions. A vector x in some set S n is mpped into m nother vector y T( x). A trnsformtion is liner if, for

More information

1.2. Linear Variable Coefficient Equations. y + b "! = a y + b " Remark: The case b = 0 and a non-constant can be solved with the same idea as above.

1.2. Linear Variable Coefficient Equations. y + b ! = a y + b  Remark: The case b = 0 and a non-constant can be solved with the same idea as above. 1 12 Liner Vrible Coefficient Equtions Section Objective(s): Review: Constnt Coefficient Equtions Solving Vrible Coefficient Equtions The Integrting Fctor Method The Bernoulli Eqution 121 Review: Constnt

More information

Reversals of Signal-Posterior Monotonicity for Any Bounded Prior

Reversals of Signal-Posterior Monotonicity for Any Bounded Prior Reversls of Signl-Posterior Monotonicity for Any Bounded Prior Christopher P. Chmbers Pul J. Hely Abstrct Pul Milgrom (The Bell Journl of Economics, 12(2): 380 391) showed tht if the strict monotone likelihood

More information

p-adic Egyptian Fractions

p-adic Egyptian Fractions p-adic Egyptin Frctions Contents 1 Introduction 1 2 Trditionl Egyptin Frctions nd Greedy Algorithm 2 3 Set-up 3 4 p-greedy Algorithm 5 5 p-egyptin Trditionl 10 6 Conclusion 1 Introduction An Egyptin frction

More information

Math 31S. Rumbos Fall Solutions to Assignment #16

Math 31S. Rumbos Fall Solutions to Assignment #16 Mth 31S. Rumbos Fll 2016 1 Solutions to Assignment #16 1. Logistic Growth 1. Suppose tht the growth of certin niml popultion is governed by the differentil eqution 1000 dn N dt = 100 N, (1) where N(t)

More information

Module 6 Value Iteration. CS 886 Sequential Decision Making and Reinforcement Learning University of Waterloo

Module 6 Value Iteration. CS 886 Sequential Decision Making and Reinforcement Learning University of Waterloo Module 6 Vlue Itertion CS 886 Sequentil Decision Mking nd Reinforcement Lerning University of Wterloo Mrkov Decision Process Definition Set of sttes: S Set of ctions (i.e., decisions): A Trnsition model:

More information

Theoretical foundations of Gaussian quadrature

Theoretical foundations of Gaussian quadrature Theoreticl foundtions of Gussin qudrture 1 Inner product vector spce Definition 1. A vector spce (or liner spce) is set V = {u, v, w,...} in which the following two opertions re defined: (A) Addition of

More information

Genetic Programming. Outline. Evolutionary Strategies. Evolutionary strategies Genetic programming Summary

Genetic Programming. Outline. Evolutionary Strategies. Evolutionary strategies Genetic programming Summary Outline Genetic Progrmming Evolutionry strtegies Genetic progrmming Summry Bsed on the mteril provided y Professor Michel Negnevitsky Evolutionry Strtegies An pproch simulting nturl evolution ws proposed

More information

13: Diffusion in 2 Energy Groups

13: Diffusion in 2 Energy Groups 3: Diffusion in Energy Groups B. Rouben McMster University Course EP 4D3/6D3 Nucler Rector Anlysis (Rector Physics) 5 Sept.-Dec. 5 September Contents We study the diffusion eqution in two energy groups

More information

We will see what is meant by standard form very shortly

We will see what is meant by standard form very shortly THEOREM: For fesible liner progrm in its stndrd form, the optimum vlue of the objective over its nonempty fesible region is () either unbounded or (b) is chievble t lest t one extreme point of the fesible

More information

different methods (left endpoint, right endpoint, midpoint, trapezoid, Simpson s).

different methods (left endpoint, right endpoint, midpoint, trapezoid, Simpson s). Mth 1A with Professor Stnkov Worksheet, Discussion #41; Wednesdy, 12/6/217 GSI nme: Roy Zho Problems 1. Write the integrl 3 dx s limit of Riemnn sums. Write it using 2 intervls using the 1 x different

More information

Line Integrals. Partitioning the Curve. Estimating the Mass

Line Integrals. Partitioning the Curve. Estimating the Mass Line Integrls Suppose we hve curve in the xy plne nd ssocite density δ(p ) = δ(x, y) t ech point on the curve. urves, of course, do not hve density or mss, but it my sometimes be convenient or useful to

More information

New Expansion and Infinite Series

New Expansion and Infinite Series Interntionl Mthemticl Forum, Vol. 9, 204, no. 22, 06-073 HIKARI Ltd, www.m-hikri.com http://dx.doi.org/0.2988/imf.204.4502 New Expnsion nd Infinite Series Diyun Zhng College of Computer Nnjing University

More information

Zero-Sum Magic Graphs and Their Null Sets

Zero-Sum Magic Graphs and Their Null Sets Zero-Sum Mgic Grphs nd Their Null Sets Ebrhim Slehi Deprtment of Mthemticl Sciences University of Nevd Ls Vegs Ls Vegs, NV 89154-4020. ebrhim.slehi@unlv.edu Abstrct For ny h N, grph G = (V, E) is sid to

More information

Linear Systems with Constant Coefficients

Linear Systems with Constant Coefficients Liner Systems with Constnt Coefficients 4-3-05 Here is system of n differentil equtions in n unknowns: x x + + n x n, x x + + n x n, x n n x + + nn x n This is constnt coefficient liner homogeneous system

More information

SCHOOL OF ENGINEERING & BUILT ENVIRONMENT. Mathematics

SCHOOL OF ENGINEERING & BUILT ENVIRONMENT. Mathematics SCHOOL OF ENGINEERING & BUIL ENVIRONMEN Mthemtics An Introduction to Mtrices Definition of Mtri Size of Mtri Rows nd Columns of Mtri Mtri Addition Sclr Multipliction of Mtri Mtri Multipliction 7 rnspose

More information

Riemann is the Mann! (But Lebesgue may besgue to differ.)

Riemann is the Mann! (But Lebesgue may besgue to differ.) Riemnn is the Mnn! (But Lebesgue my besgue to differ.) Leo Livshits My 2, 2008 1 For finite intervls in R We hve seen in clss tht every continuous function f : [, b] R hs the property tht for every ɛ >

More information

Review of basic calculus

Review of basic calculus Review of bsic clculus This brief review reclls some of the most importnt concepts, definitions, nd theorems from bsic clculus. It is not intended to tech bsic clculus from scrtch. If ny of the items below

More information

Extended nonlocal games from quantum-classical games

Extended nonlocal games from quantum-classical games Extended nonlocl gmes from quntum-clssicl gmes Theory Seminr incent Russo niversity of Wterloo October 17, 2016 Outline Extended nonlocl gmes nd quntum-clssicl gmes Entngled vlues nd the dimension of entnglement

More information

P 3 (x) = f(0) + f (0)x + f (0) 2. x 2 + f (0) . In the problem set, you are asked to show, in general, the n th order term is a n = f (n) (0)

P 3 (x) = f(0) + f (0)x + f (0) 2. x 2 + f (0) . In the problem set, you are asked to show, in general, the n th order term is a n = f (n) (0) 1 Tylor polynomils In Section 3.5, we discussed how to pproximte function f(x) round point in terms of its first derivtive f (x) evluted t, tht is using the liner pproximtion f() + f ()(x ). We clled this

More information

CS 373, Spring Solutions to Mock midterm 1 (Based on first midterm in CS 273, Fall 2008.)

CS 373, Spring Solutions to Mock midterm 1 (Based on first midterm in CS 273, Fall 2008.) CS 373, Spring 29. Solutions to Mock midterm (sed on first midterm in CS 273, Fll 28.) Prolem : Short nswer (8 points) The nswers to these prolems should e short nd not complicted. () If n NF M ccepts

More information

PHYS Summer Professor Caillault Homework Solutions. Chapter 2

PHYS Summer Professor Caillault Homework Solutions. Chapter 2 PHYS 1111 - Summer 2007 - Professor Cillult Homework Solutions Chpter 2 5. Picture the Problem: The runner moves long the ovl trck. Strtegy: The distnce is the totl length of trvel, nd the displcement

More information

Chapter 5. Numerical Integration

Chapter 5. Numerical Integration Chpter 5. Numericl Integrtion These re just summries of the lecture notes, nd few detils re included. Most of wht we include here is to be found in more detil in Anton. 5. Remrk. There re two topics with

More information

Lecture 3 Gaussian Probability Distribution

Lecture 3 Gaussian Probability Distribution Introduction Lecture 3 Gussin Probbility Distribution Gussin probbility distribution is perhps the most used distribution in ll of science. lso clled bell shped curve or norml distribution Unlike the binomil

More information