ࡱ> Y[X5@ -Sbjbj22 ֞XXJ; T T T T VVVDVuuu>vwTV'rwny~y~y~yLJ($RQRV^"T"T~y~yN"T~yV~y >8UV~yfw pEMbuv80'ٖpVV"T"T"T"TVrrrrrVVD:j c"VV:jRatings in Distributed Systems: A Bayesian Approach Lik Mui, Mojdeh Mohtashemi, Cheewee Ang, Peter Szolovits MIT Laboratory for Computer Science 200 Technology Square, Cambridge, MA 02139 {lmui, mojdeh, cheewang, psz}@lcs.mit.edu Ari Halberstadt Magiccookie 9 Whittemore Road, Newton, MA 02458 ari@magiccookie.com Abstract For distributed systems at large and e-commerce systems in particular, ratings play an increasingly important role. Ratings confer reputation measures about sources. This paper reports our formalization of the rating process. This paper argues that ratings should be context- and individual- dependent quantities. In contrast to existing rating systems in many e-commerce or developer sites, our approach makes use of personalized and contextualized ratings for assessing source reputation. Our approach is based on a Bayesian probabilistic framework. Introduction In a networked environment where resources are distributed widely, locating relevant resources has been the subject of the information retrieval industry. However, reliability measures for the retrieved data are often missing. This paper reports our progress on the development of a mathematical scheme and computational tools for how approval ratings can be propagated and assimilated for the purpose of assigning reputation values to other agents. Degrees of approval are modeled in a probabilistic framework. We adopt the following definition for reputation: Reputation: perception that an agent has of anothers intentions and norms. Discussion on the meaning of this definition can be found in [Mui, et al., 2002], along with how reputation relates to concepts such as trust: Trust: a subjective expectation an agent has about anothers future behavior based on the history of their encounters. We argue that the failure of existing rating systems is partially due to the lack of personalization and contextualization in these systems. For example, consider an expert in French cuisine browsing the Amazon.com catalogue, she comes across a French cookbook with a 3 stars rating (as provided by other raters). The average Amazon user is not likely to be a French cuisine expert. For our expert, the three star rating is therefore of limited value. Unless the rating system can summarize the ratings of those who are similar to the individual seeking a rating, the rating system is destined to provide noisy ratings. This example points to the importance of personalizing ratings for those seeking source sanctioning. Similarly, a users reputation as a computer scientist is likely to be different from his or her reputation as a violinist. Existing systems such as those in eBay and Slashdot do not take into account this context dependency. Section 2 covers the relevant literature. Section 3 formalizes intuitions concerning ratings. Section 4 concludes this paper with thoughts on future works. Background The information retrieval (IR) community has been tackling the question of information relevance for decades now, and with remarkable success (Salton, 1971; Kleinberg, 1997; Brin and Page, 1998). However, simply locating relevant resources in distributed systems today is no longer adequate. Users demand not only relevant but also reliable information. Collaborative filtering is a paradigm for using the collective experiences of a cyber community to select relevant resources (Goldberg, et al., 1992). Large numbers of users are often required before reasonable performance is achieved. No metric exists to measure how reliable is a rating nor how reputed is a rater. The recent human-computer interface (HCI)s interest in social navigation highlights the importance of leveraging collective experience in aiding users interacting with information systems (Dieberger, et al., 2000; Wexelblat, 2001). However, techniques for assessing the reliability of the collected data are ad-hoc. Prompted by the rise of on-line communities and their proliferation in the past decade, rating systems have been built to provide source reputation or reliability assessment. On-line auction eBay provides a +1/-1 rating system for buyers to provide feedback about sellers and vice versa. A few recent high profile fraud cases in eBay by individuals with high eBay ratings suggest that the company should seriously consider enhancing their simple rating system. Rating systems in such commercial sites aggregate all users ratings equally and ignore the personalized nature of reputation (i.e., ones reputation means different things to different individuals depending on the context). We term this type of reputation measure global in nature as opposed to being personalized for the inquirers context. In the computer science literature, Zacharia and Maes (1999) have suggested that reputation in an on-line community can be related to the ratings that an agent receives from others, and have pointed out several criteria for such rating systems. However, their mathematical formulation for the calculation of reputation can at best be described as intuitive without justifications except the intuitive appeal of the resulting reputation dynamics. Ad-hoc formulation plagues several other proposals for rating or reputation systems such as those in Glass, et al. (2000), Yu and Singh (2000), Rouchier, et al. (2001), Sabater, et al. (2001), Esfandiari, et al. (2001), among others. Nevertheless, ratings and reputation measures have been found to provide useful intuition or services for users of these systems. Conceptual Framework On notations, consider the universe of users and objects in the system as the following sets: Set of agents: A = { a1, a2, aM } Set of objects: O = { o1, o2, oN } Users rate the objects, which could be other users (i.e., A ( O). The set of ratings ( can be viewed as a mapping from A x O to a real number between 0 and 1 (to accommodate probability). Rating:  EMBED Equation.DSMT4  To model the process of opinion sharing among agents, the concept of an encounter is required. An encounter is an event between two different agents (ai, aj) such that the query agent ai asks the responding agent aj for ajs rating of an object: Encounter:  EMBED Equation.DSMT4  where {(} represents the set of no encounter (bottom). Consider an agent ai who has never interacted with object ok in the past. Before taking the trouble of interacting with ok, ai asks other agents (A-i) in the environment their ratings for ok. ai will decide to interact with object ok, for example, if the weighted sum of the ratings (a proxy for oks reputation) from agents in A-i is greater than a certain threshold. The reputation of aj in ais mind can be considered as the probability that in the next encounter, ai will approve of ajs rating for the object context. A context describes a set of attributes about an environment. Let an attribute be defined as the presence (1) or absence (0) of a trait. The set of all attributes is possibly countably infinite, and is defined as follows: Attribute:  EMBED Equation.DSMT4  Set of attributes: B = { b1, b2, } A context is then an element of the power set of B: Context: c ( C and C = P{B} where P{.} represents the power set. The reputation mapping can now be represented by: Reputation:  EMBED Equation.DSMT4  This mapping assumes that reputation only changes after an encounter either directly with the individual involved, or through rating feedbacks from other agents. After each encounter, the update rule for the state of the system can be represented as below (which is discussed in the next section): Update rule: newstate: R D ( R Bayesian Reputation Inference Let xab(i) be the indicator variable for agent as approval of b after the ith encounter between them thus the ratings ( is restricted to binary values of 0 or 1. If a and b have had n encounters in the past for context c, the proportion estimator  EMBED Equation.DSMT4  for the number of approvals of b by a can be modeled with a Beta prior distribution (Dudewicz and Mishra, 1988):  EMBED Equation.DSMT4  where c1 and c2 are parameters determined by prior assumptions and  EMBED Equation.DSMT4  is the estimator for the true proportion of as approval of b based on encounters between them for context c. (From this point on, a given context c is assumed and the subscript is avoided for clarity.) Assuming that each encounters approval probability is independent of other encounters between a and b, the likelihood of having p approvals and (n p) disapprovals can be modeled as:  EMBED Equation.DSMT4  Combining the prior and likelihood, the posterior estimate can be shown to be (Mui, et al., 2001):  EMBED Equation.DSMT4  In our framework, reputation for b in as mind is as estimate of the probability that a will approve of b in the next encounter x(n + 1). This estimate is based on n previous encounters between them (for the specific context c). This estimate can be shown to yield the following (ibid.):  EMBED Equation.DSMT4  (1) This conditional expectation is the expression for rab : the update rule specified in the last section. First order statistical properties of the posterior are summarized below for the posterior estimate of  EMBED Equation.DSMT4 :  EMBED Equation.DSMT4  which indicates an easy implementation for the estimation procedure. Note that rab in (1) is a personalized estimate by agent a, which depends on the set of encounters D that agent a takes into account about b. Specifying the set of encounters is individual dependent and could include inferred encounters from hearsays from agent as neighbors, as suggested in the propagation procedures below. We also have not specified how prior belief about  EMBED Equation.DSMT4  is formed? Two situations are considered below. Complete Strangers: Prior Assumptions If agents ai and aj are complete strangers, an ignorance assumption is made. When these two strangers first meet, their estimate for each others reputation should be uniformly distributed across the reputations domain:  EMBED Equation.DSMT4  For the Beta prior, values of c1=1 and c2=1 yields such a uniform distribution. Known Strangers: Prior Assumptions If ai and ak have never met before but ai knows aj well (i.e., ai has an opinion on ajs reputation). Also, aj knows ak well. ai would like to estimate aks reputation based on ai s history of encounters with aj, and ajs history of encounters with ak. This setup is depicted below:  Figure  SEQ Figure \* ARABIC 1. Illustration of an indirect neighbor pair (ai, ak) For agent ai, agent ak is a stranger but ak is not completely unknown to ai since ai knows aj who has an opinion about k. In a future encounter between ai and ak, the probability that this encounter will be rated good by ai is: Let Dij,n = all n encounters between ai and aj. Djk,m = all m encounters between aj and ak. xij(n) = indicator variable for ais approval of ak at encounter n  EMBED Equation.DSMT4  (2) The interpretation of this equation is that the probability that ai would approve ak at encounter n + 1 is the sum of the probabilities that both ai and aj agree (about the context that c is in) and that both of them disagree. Inference Propagation So far, we have discussed belief estimation for indirect neighbors which are one degree away from a direct neighbor. The reputation of indirect neighbors which are further away can be estimated by recursively applying Equation (2) using closer neighbors as direct neighbors. This procedure works if there exists a single path between two agents. This recursion has a fixed point for a finite population community: when all members of the communities have been rated, the recursion stops. If there are multiple paths between two agents, as illustrated in  REF _Ref527259914 \h  \* MERGEFORMAT Figure 2, a different procedure is required.  SHAPE \* MERGEFORMAT  Figure  SEQ Figure \* ARABIC 2. Illustration of a parallel network between two agents a and b.  REF _Ref527259914 \h  \* MERGEFORMAT Figure 2 shows a parallel network of k chains between two agents of interest, where each chain consists of at least one link. Agent a would like to estimate agent bs reputation as defined by the embedded network between them. Clearly, to sensibly combine the evidence about b through all intermediate agents in between, measures of reliability about these intermediates are required. One way these measures can be calculated is shown in Mui, et al. (2002) by using Chernoff bound on the parameter estimate  EMBED Equation.DSMT4 . Let m represents the minimum number of encounters necessary to achieve a given level of confidence ( for  EMBED Equation.DSMT4  and error bound (; the following inequality can be demonstrated:  EMBED Equation.DSMT4  As ( approaches 1, a larger m is required to achieve a given level of error bound (. This threshold (m) can be set on the number of encounters between agents such that a reliable measure can be established as follows:  EMBED Equation.DSMT4  where mab is the number of encounters between agents a and b. Intuition of this formula is as follows: m represents the minimum sample size of encounters to reach a confidence (and error) level about the parameter estimates. Above a given level of sample size, the estimator is guaranteed to yield the specified level of confidence. Therefore, such an estimate can be considered as reliable with respect to the confidence specification. Any sample size less than the threshold m is bound to yield less reliable estimates. As a first order approximation, a linear drop-off in reliability is assumed here. For each chain in the parallel network, how should the total weight be tallied? Two possible methods are plausible: additive and multiplicative. The problem with additive weight is that if the chain is broken by a highly unreliable link, the effect of that unreliability is local to the immediate agents around it. In a long social chain however, an unreliability chain is certain to cast serious doubt on the reliability of any estimate taken from the chain as a whole. On the other hand, a multiplicative weighting has long-distance effect in that an unreliable link affects any estimate based on a path crossing that link. The form of a multiplicative estimate for chain is weight (wi) can be:  EMBED Equation.DSMT4  where li refers to the total number of edges in chain i and wij refers to the jth segment of the ith chain. Once the weights of all chains of the parallel network between the two end nodes are calculated, the estimate across the whole parallel network can be sensibly expressed as a weighted sum across all the chains:  EMBED Equation.DSMT4  where rab(i) is as estimate of bs reputation using path i and  EMBED Equation.DSMT4  is the normalized weight of path i ( EMBED Equation.DSMT4 sum over all i yields 1). rab can be interpreted as the overall perception that a garnered about b using all paths connecting the two. Discussion and Conclusion This paper has briefly reported our Bayesian formalization of the rating process. We have argued that since ratings and reputations are clearly context dependent quantities, models about them should explicitly take the context into account. Applications of this formalization have been applied to simulated and real world ratings and are reported in Mui, et al., 2001. Our Bayesian approach achieves more accurate reliability and reputation estimates than control schemes similar to those used in existing systems such as eBay and Amazon. There are several yet to be resolved issues. We have examined binary ratings in this paper. Extending the Bayesian framework to other types of ratings should be straightforward. Specifying the procedure for inferring ratings and reputations from one context to another is also needed. We have started investigating the use of ontologies to relate different contexts. A hard problem is to determine how to resolve the different ontological views of the world held by different agents. Furthermore, the metric or function to transfer rating or reputation from one context to the next is yet to be worked out. In our formulation, the calculation of an agent's reputation requires the disclosure of detailed personal rating and object descriptions to other agents. This creates a privacy concern which needs to be addressed. Acknowledgements We thank the support of the MEDG group in MIT/LCS for providing a supportive research environment. We especially thank Dr. Jon Doyle for his discussions. References S. Brin, L. Page (1998) "The Anatomy of a Large-Scale Hypertextual Web Search Engine," Proc. 1998 WWW Conference. A. Dieberger, P. Dourish, K. Hook, P. Resnick, and A. Wexelblat (2000) Social Navigation: Techniques for Building More Usable Systems, Interaction, 7(6), pp. 36-45. E. J. Dudewicz, S. N. Mishra (1998) Modern Mathematical Statistics, John Wiley & Sons. B. Esfandiari, S. Chandrasekharan (2001) On How Agents Make Friends: Mechanisms for Trust Acquisition, 4th Workshop on Deception, Fraud and Trust in Agent Societies, Montreal, Canada. A. Glass, B. Grosz (2000) Socially Conscious Decision-Making, Autonomous Agents2000. D. Goldberg, D. Nichols, B. M. Oki, D. Terry (1992) Using collaborative filtering to weave an information tapestry. Communication of the ACM, 35(12), pp. 61-70. J. Kleinberg (1997) Authoritative sources in a hyperlinked environment. IBM Technical Report RJ 10076. L. Mui, M. Mohtashemi, C. Ang, P. Szolovits, A. Halberstadt (2001) Bayesian Ratings in Distributed Systems: Theories, Models, and Simulations, MIT LCS Memorandum. L. Mui, M. Mohtashemi, A. Halberstadt (2002) A Computational Model of Trust and Reputation, Proc. 34th Hawaii International Conference on System Sciences. J. Rouchier, M. OConnor, F. Bousquet (2001) The Creation of a Reputation in an Artificial Society Organized by a Gift System, J. Artificial Societies and Social Simulations, 4(2). J. Sabater, C. Sierra (2001) REGRET: A reputation Model for Gregarious Societies, 4th Workshop on Deception, Fraud and Trust in Agent Societies, Montreal, Canada. G. Salton (1971) SMART Retrieval System: Experiments in Automatic Document Processing, Englewood Cliffs, NJ: Prentice-Hall. S. Wasserman, K. Faust (1994) Social Network Analysis: Methods and Applications, Cambridge University Press. A. Wexelblat (2001) History-Rich Tools for Social Navigation, MIT Media Lab Report. B. Yu, M. P. Singh (2000) A Social Mechanism of Reputation Management in Electronic Communities, Cooperative Information Agents2000. G. Zacharia, P. Maes (1999) Collaborative Reputation Mechanisms in Electronic Marketplaces, Proc. 32nd Hawaii International Conference on System Sciences.  Although different in other contexts, reliability here is taken to confer source reputation. The intuition used here is that the more reputed is a source, the more reliable is information gathered from it. PAGE  b Chain k Chain 2 a Chain 1 ai ak aj  34o< ? G H + 5 S T \ s t u  ' : ; @ ŲŲŠysmga]YU]QMIhF&h+H>hwu2hah2C*h`H hU ? H u gd(ogd|/ `gd(o $x`a$gd(o$]`a$gd?4] $`a$gd(o $`a$gd(o `gd(oRRR,S@ S V e f  E HNSTU GJfsno%+.n0v⹵⫵ hwu2hwu2hwu2OJQJh hqSh hFh hoh 56h hXf/h='hbb6hV hFjD hFhbbhohbb56h1hbbh }hajh[Z0JUh+H>h`H6  cnS$n,Qz$ @ xxa$gd` 0]xx^]`gd`gddgdigd|/`gd & F&  xx^` gd `gdbb & F&  xx^` gdbbv|3[a6<KL]^| GKQRS~67;Bh?DEFGHV_`adejkl +,2456MNOPy~ 쾴쾬안hd}jhdh\kEHU-jΉ? h\kCJOJQJUVaJnHtHjh`HUhn=@OJQJ^Jh&hg jrh+h`H6 jh05H h05H6h05H h`H5\hJh`H h`H6]h`H6H*]4 ' ( ) . / 0 H Q S T U l m n o p w x "!#!$!&!'!(!=!?!f!g!h!k!l!m!!!!!!!}hdhd6hdha66H*] ha66]ha6 h:vbh  j^h mHsHh j/hdh\kEHU-jΉ? h\kCJOJQJUVaJnHtHjh`HUh& h`H5\h`H6H*] h`H6]h`Hhh0QH p ##$=$$$%&-&' @>0^`0gd)Q$ @ xxa$gdkw `gdL`gdiQ ` @ xxgdgdK`gd $ @ xxa$gd 0` !!!!!!","/"0"1"5"6"7"""""""""""#####################$&$($*$+$}h ah a6 h`HH*jhiQhU)EHU-j/? hU)CJOJQJUVaJnHtHjh`HU h`H5\hiQh\Ah\A6H*h\Ah\A6h\A h`H6]h\Ah`H5h`H6H*]h`Hhdhdhd6H*0+$,$-$8$9$<$C$D$a$c$$$$$$$$$$$$$%?%U%z%{%%%%%%%%%&&&&& & & & & &&鰣闏~zs jh+EhU) hU)6h+Eh+E6h+Ehuh+E6hKha6hV5j hQ:h\kEHU-j ω? h\kCJOJQJUVaJnHtHjh`HUh8OshQ: h`H5\hiQ h`H6]h`Hh ah a6 jh ah a-&&&#&1&2&4&5&6&V&\&]&l&m&x&y&{&&&&&&&&&&&& ' ''(')'@'A'B'C'L'c'd'h'i''''''ܿ{jh`HUj h5h5EHU-j? h5CJOJQJUVaJnHtHjh4kUh4kh5h56h5 jrhtht6h4kh`H6H*h4kh`H6htha6h`H6H*ha6h`H6h)Qh`H h`H6.''''''''''(((,(-(.(/(\(](l(m((((((((((\)])b)c)~)))))ɿɿѻ|tojjbh9h`H6 h`H6 h56h5h56 hvhzhvhz6jh5h5EHU-j? h5CJOJQJUVaJnHtHh5jh5UhzhEh`H6H*hEh`H6h`Hjh`HUjh5h5EHU)j? h5CJOJQJUVnHtH&''))5*Q*u++,,..//0|`gdkw !`gdkwgdkwgdJ`gdz $`a$gdz`gdz !$`gdz $ pa$gdB`gdu{7$a$gdB`$ :a$gdB))))))*$*&*,*3*5*6*M*N*O*P*r*s*w*x*************+4+5+6+l+q+u+v+w+++󏇏n)j%ω? h\kCJOJQJUVnHtHhQ:h5h56h5 h`H6] h`H6jhzh\kEHU)j ω? h\kCJOJQJUVnHtHh)Qh)Q6h)Qh^pjyh9h\kEHU)jω? h\kCJOJQJUVnHtHh`Hjh`HU++++++++++++++f,g,~,,,,,,,,,,,,,,,,,,ļĎwjffbZhh6hha6j%hBHhYEHU-j? hYCJOJQJUVaJnHtHjhzUj"h ahtEHU)j? htCJOJQJUVnHtHhtjhtUhzhQhP1h`H6H*hP1h`H6hP1h`HhQ:jh`HUjh\kh\kEHU ,,,,----H-I-U-V-p-q---M.`.a.x.y.z.{.|.............R/y//////һާޣޜޜސވwmja-h1EHU!j > h1CJOJQJUVjh1UhJh16H*] h16]ha6 hJh1j*h ahtEHU-j? htCJOJQJUVaJnHtHhtjhtUh1hh6hKh2hhh6H**////////0"0C0D0E0J0K0L0g0h0i0j0p0q0r0y000000000000000000000000111111<1=1>1_1`1a1f1m1n1²jh15U\ h15\ hz]jhbzUmHnHtHu h1] h16 h16H* h16] hJh1hD@h16H*hD@h16h1jh1U=0@0_1a1b1c1d1e1f111223H3i3M4c4N667 $$`a$gdWgd$ !$a$gdz !8^`gd1$a$gd1$a$gd1gd1gdJn111111111111111111122 22222225262W2X2Y2Z2^2_2`22222222222222222222222222333333 3 3 h`[F6 h16H* h16 h1]hfLh16H*] h16] h1H*h1hAJ5\mHnHujh15U\ h15\E 3 3%3&3'3637383F3H3I3J3a3b3c3d333333333333444#4$4M4W4b4c4r4}4K5Q5x5555N6~666깱깱깱깱ꭦ~jh2;hfLUaJ hfLh1ha6hfLh oh,\h$mC hJh`Hh`Hh16H*] h16]j1h1h1EHU)j? h1CJOJQJUVnHtHjh1Uh1 h16H* h1] h16/6666666666666777777 7 7#7$7%7&7_7`7a7d7e7ʽԴ~teteeWtKWKh2;h0J6aJh2;h0J56aJjh2;h6UaJh2;h6aJj}7hUjhUmHnHujhUjh%IRUmHnHtHuhhOZ*h2;hOZ*aJhAJ6aJmHnHuh2;hAJ6aJjh2;hfLUaJ h2;aJj7h2;h*UaJh2;hfLaJe7f7g7h7i7777777777886878888888 999R9X99999ŵΥΔ|w|o|`jl? h|UUVnHtHjh|UU h|U6h|UhUh6hih%h6hhAJ6aJmHnHuh2;hAJ6aJ h2;aJj 8h2;hUaJh2;haJjh2;hUaJh2;h0J56aJh2;h0J6aJh2;h6aJ 7h7v::m;;=@@7A B&BJCdCEHHmIxI`gd|/gd|/$a$gdgd`gd $xa$gdJUL`gdH) $xxa$gd2;gd|U $a$gd+99999::::1:2:3:4:E:F:h:t:u:v:w:::::::::::::::Ǻ毫|tla]h jeh.Wh|U6h*bh|U6 jdh|U6j>h hYEHU-jZ? hYCJOJQJUVaJnHtH hoh|UhH)h2; jeh|Uh|U6jH;h oh|UEHUjl? h|UUVnHtH jdh|Uh|U6 h|U6h|Ujh|UUj8h oh|UEHU!::=;F;m;n;;;;;;;;;;;;;;;H<i<l=m=@@@@@@@@@@@@@@AArejFhYhYEHU-jc? hYCJOJQJUVaJnHtHh"h6H*h"h6 ht 6ht ht ht 6 h6h yh6H*h yh6jAh[?yhYEHU-jU? hYCJOJQJUVaJnHtHjhUh|Uhhi/2h6'AAA AAAA,A-A/A B B"B#B$B%B&B+B,B-B.B/B0B1B6B7BFBGBHBJB`BaBfBgB~BBBBBBBB~qihh6jMhdhEHUjB'm? hUVaJha6 h6 h6H*hZh6H*hZh6 hZhjIhYhYEHU-jj? hYCJOJQJUVaJnHtHjhUhK=h6H*hK=h6H*hK=h6h)BBBBBBBBBBBCC#C$CJCYCsC{CCCCWDDDDwEEEEEE9FHFRFFFFFFFmGnGGHHHHž}hlDhRhBh4hkh }hqXhqpXh$ ha66ha6hCh:"h\h`Hht h6H*ht=h6 h6hh6jhUjPhdhEHUjB'm? hUVaJh/HH6IDIJINI`IlImIwIxIIIIIIII J)J0JsJJJJJJJJJJSKKKKKKKKKKֽ~zvnfhp"hp"6hp"h`H6hfhp"h3h36mHsHh3h3mHsH hf6hP h`H6hP mHsHhTh`HmHsHhfhqo6hqo hqohqo hP 6] h`H6]hMhUE3 3t UDn`!M (>E3 3t J pxuRkAl6@W1YEE6z&z06.HD"ZzQ< ңG"''%M0&Qb]|ޛC_En9:Ӎ_D|!NDW:#kڌ#yEwJͻPPQ2hHM~׿T:xFjUxɫ춗`I?y@ v;242g.,fO_|$_YQKhH7_ xenǬGMpLW.7{!yAzeb|i4~bV*^r mxpqOh_ <4!?cႄύM/}-TsmFۓ/pYMaw&1{9fTIԟηdkoڨpDd hb  c $A? ?3"`?2|H#RK"sn`!|H#RK" @(|\xڝSMhA~v7&i -Ji~@RmI `l0)=6]͢$%PQKBEcЫIOJXTov7)Atؙޛ߼7 Dph씥‘ #A&t9rcLW4;$|pS jn~OHQ<rtu.;nj AZwCn65/M_m "] kb!'w⮡f ;/6353\\hkc^cx?,p Lfg&G9H׋JEH :P< wgpXXT٨Ld2R(DJvRiG#J;V_~q jVI JJ?V@Ogv4KHvhx}' >I:&t)n;aHni Bx[!8aT`[I0w%:?\e.nXZfclGo0׍ښ!1^% 땪V4{O Gnո1BDd < ,b  c $A? ?3"`?2`Y|[Yk ;hn`!``Y|[Yk ;~`Hb.xڕSKkQ$ LZE)f"T<`RL'f<$Ph *n"n+K`VR\ԥ[Auو`<6I!{y3/xK@qD}D)tN11O`tέa = 7tᥳݍHL'$:."(/t!JVMe5ەYƉ^#Koa8Y"k|.Wl~i=<3w#|@g/xtxΟl J9a͹]\&nr"&2߫wƓ vi+I.B9Z,ZNR]kڠOܯ3UYe.fnL4eOjQ ݊e8ݐ3`Va[̟.d|$ႈ t@5׋*p>L lX&C5/E'K|:3Cqn6dOUN5ᦺ ި?SN7ku$% ې+"Db<,91SryF"qX#3.o,T= Tk;](Q2~3@ZohVsIhG"fgzt[y0̍lfy"FZc|;zE#+ɬ^F's[;+'{ pKpvPkbΟGJӵx:^k#HwLzE~?=)EDd |h  s *A? ?3"`?2=N.O 0n@ n`!N.O 0n `R0xڕRAkQm7m`7V Ҧ "^e,`JhR{og&K-D"cnJҢ&4唉9)X|3`w9b*J^U>PS,OHtS<=sU>2Vu?n[6f $b7wG p/#n7)BzxDtB$8G%lh7 栱;nnnBe+cƶln^hoeEF/%dBMwOd&JլpW~q.GԵg o6"'.#,Hww-ꫛwk :l3M ^,q&ߴCC,C{WOz)mJ[NGj.9R%<X4;{`55@Dd  !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOQRSTUVWZ]{^_`abcdefghijkmlonpqrstuvwxzy|}~Root Entryi F]Mb\<Data PySWordDocumenth֞ObjectPoolk'08Mb]Mb_1065995988cF08Mb08MbOle CompObjiObjInfo "#$%(+,-034567:=>?BEFGHIJKLMNQTUVWXY\_`abcdgjklorstuvwxyz{|}~ FMathType 4.0 Equation MathType EFEquation.DSMT49q @6M6GDSMT4WinAllBasicCodePagesTimes New RomanSymbolCourier NewMT Extra!/'_!!/G_APAPAE%B_AC_A %!AHA_D_E_E_A  @ `r:AEquation Native (_1065996029J FоMbоMbOle  CompObj iO![0,1] FMathType 4.0 Equation MathType EFEquation.DSMT49qU@6M6GDSMT4WinAllBasicCodePagesTimes New RomanSymbolCourier NewMT Extra!/'_!!/G_APAPAE%B_AC_A %!AHA_D_E_E_ObjInfo Equation Native  q_10660022231OFоMbоMbOle A  @ `d"D==A @ `2 `O[0,1]*"{"^} FMathType 4.0 Equation MathType EFEquation.DSMT49q3@6M6GDSMT4WinAllBasicCodePagesCompObjiObjInfoEquation Native O_1065996041FоMbоMbTimes New RomanSymbolCourier NewMT Extra!/'_!!/G_APAPAE%B_AC_A %!AHA_D_E_E_A  @ `b"Bandb:O!{0,1} FMathType 4.0 Equation MathType EFEquation.DSMT49qOle CompObjiObjInfo Equation Native !< @6M6G4DSMT4WinAllBasicCodePagesTimes New RomanSymbolCourier NewMT Extra!/'_!!/G_APAPAE%B_AC_A %!AHA_D_E_E_A  @ `R:AAC![0,1] FMathType 4.0 Equation MathTy_1066046085FоMbоMbOle &CompObj'iObjInfo)pe EFEquation.DSMT49q@6M6GDSMT4WinAllBasicCodePagesTimes New RomanSymbolCourier NewMT Extra!/ED/APG_APAPAE%B_AC_AE*_HA@AHA*_D_E_E_A   `q  c FMathType 4.0 Equation MathTyEquation Native *_1066046102"FоMbоMbOle .CompObj /ipe EFEquation.DSMT49qY@6M6GDSMT4WinAllBasicCodePagesTimes New RomanSymbolCourier NewMT Extra!/ED/APG_APAPAE%B_AC_AE*_HA@AHA*_D_E_E_A   `p(q  c `)==Beta(c  `1 `,c  `2  `)ObjInfo!1Equation Native 2u_1066046112$FоMbоMbOle 8 FMathType 4.0 Equation MathType EFEquation.DSMT49q@6M6GDSMT4WinAllBasicCodePagesTimes New RomanSymbolCourier NewMT Extra!/ED/APG_APAPAE%B_AC_AE*_HA@AHA*_D_E_E_A   `q  cCompObj#%9iObjInfo&;Equation Native <_1065996055 6)FоMbоMbOle @CompObj(*AiObjInfo+CEquation Native D FMathType 4.0 Equation MathType EFEquation.DSMT49qh@6M6GDSMT4WinAllBasicCodePagesTimes New RomanSymbolCourier NewMT Extra!/ED/APG_APAPAE%B_AC_AE*_HA@AHA*_D_E_E_A  @ `L(D|q )==q @ `p `(1"-q ) @ `n"-p  where  `D=={x @ `ab `(1),x @ `ab `(2),...x @ `ab `(n)}_1065996064.FоMbоMbOle OCompObj-/PiObjInfo0R FMathType 4.0 Equation MathType EFEquation.DSMT49qq@6M6GDSMT4WinAllBasicCodePagesTimes New RomanSymbolCourier NewMT Extra!/ED/APG_APAPAE%B_AC_AE*_HA@AHA*_D_E_E_A  `p(q |D) == `Beta(c `1Equation Native S_1065996069,E3FоMbоMbOle ZCompObj24[i `++p,c `2 `++n"-p) FMathType 4.0 Equation MathType EFEquation.DSMT49q@6M6GDSMT4WinAllBasicCodePagesTimes New RomanSymbolCourier NewMT Extra!/ED/APObjInfo5]Equation Native ^_10660428038FоMbоMbOle eG_APAPAE%B_AC_AE*_HA@AHA*_D_E_E_A  @ px @ab @(n++1)==1|D()==Eq  @n @|D[] `==r @ `ab FMathType 4.0 Equation MathType EFEquation.DSMT49qCompObj79fiObjInfo:hEquation Native i_1066041472=FоMbоMb@6M6GDSMT4WinAllBasicCodePagesTimes New RomanSymbolCourier NewMT Extra!/ED/APG_APAPAE%B_AC_AE*_HA@AHA*_D_E_E_A   `q  FMathType 4.0 Equation MathType EFEquation.DSMT49qOle mCompObj<>niObjInfo?pEquation Native qP4@6M6GDSMT4WinAllBasicCodePagesTimes New RomanSymbolCourier NewMT Extra!/ED/APG_APAPAE%B_AC_AE*_HA@AHA*_D_E_E_A   `E  `q |D[]==  `c  `1 `++pc  `1 `++c  `2 `++ns  `q |D  `2 `==  `(c  `1 `++p)(c  `2 `++n"-p)(c  `1 `++c  `2 `++n"-1)(c  `1 `++c  `2 `++n)  `2_1049824168BFоMbоMbOle CompObjACiObjInfoD FMathType 4.0 Equation MathType EFEquation.DSMT49qOs@6M6GDSMT4WinAllBasicCodePagesTimes New RomanSymbolCourier NewMT Extra!/ED/APG_APAPAE%B_AC_AE*_HA@AHA*_D_E_E_A   `p(q )==  `10<<q .o\סdmnZ[im( V:g4` !ZҚL[52ydz.:H ֧*fؾ?w{ب*8?,V8nF!\/∳YL(ˇᤐh/edLO 1$"ܴ;oo}zjl䛊qN>{+{H{EDd |h  s *A? ?3"`?2=N.O 0nn`!N.O 0n `R0xڕRAkQm7m`7V Ҧ "^e,`JhR{og&K-D"cnJҢ&4唉9)X|3`w9b*J^U>PS,OHtS<=sU>2Vu?n[6f $b7wG p/#n7)BzxDtB$8G%lh7 栱;nnnBe+cƶln^hoeEF/%dBMwOd&JլpW~q.GԵg o6"'.#,Hww-ꫛwk :l3M ^,q&ߴCC,C{WOz)mJ[NGj.9R%<X4;{`55@\Dd $|h  s *A? ?3"`?2y % b !a|n`!ty % b !ah&` ;0BxڥUOQmD%|%bKHK1BDC[t#$@[ b1C E&Ƅp4fonJb*;fߛ̼/3 8cy\.J.v,Wu=+jZL6!'az$' dQײb0/@Pbjx7(Ցo@}QQxvtX5ɉ PG DpRk'm+]aDi!Ncy VY* 9N"QrediOyY>upkjބ2->ONX->VT=79~8J H&Фxxf7ŵܪ eD^o2`^JwgJC]9l𝘒2,!YȒ [6eˋg m,k\@.vn<`͓ҲiY }rq!attEpޑAt=3 Ŧ/^CW*W)c4^NQ1JàїGCn4u2b 6ԓ\ksC)''@'Oi +$[ހOfeT 顒0RZU3zgX1XєتWHKt|+͛H~[=$sIl3 d7'JKu&UE~芺,#wbFbfXh$!4jՃDjyI (BMaEi"Vm>۔k+ɕ(J5K-c$/ukkޮ.;ΣYnY c~3J+I8`@Dd hh  s *A? ?3"`?2vCJ㪻Zon`!vCJ㪻Zo^ @|xڥTMLQmڥI[PHh(ƃ&пALcK) d{?Od{_ܟw lHe'%O8dpsX zz|Qی7-F_D̨rX7S 9DhGP]" Azp48K_^9YZۥ;8jv{,[ pxlRG[6c(+=O{055K .̜љ^FJiTPe@q ҝ1'pMf{98;l$F)ﻘ~`|4OXC#if'̆%`6MyAL ,2/= 2%.pWOH&\O0&,CvU-9v}Ӵ(PVshvQP-agp*:{$_-%k}5`to_zR5 rϨVkOb9^%Dd W h  s *A? ?3"`? 2i'ޝvj BEn`!='ޝvj Bh xڥUMLSAV)%}ĈR4*(-7rmDBJUS@bLjC3+L0@Rxi[^~3_雈)e.Țe#iJIA|c}hwiT('~y+؀1]X> c tT0 %b2X8mXo'b8c-zè5 _H.w=XUCeʤ!`(ۯ:$RQCb)9T38<, &g ].I5UxmW}]B #фU=Z͇<^^=ޏG}% 켮-Di2P=U^癜t݉ĝѨ+̹¾p,ղHJ4;( Xt~wq!N8?{lp~"섢xwHnx1/4jIp<oN 2z(H$Ym2hV\)t.|!A Q.xb V"LtPv7]6 V 0Ҽ!kjL {VsfpkoC~g"Qަg)?'pP2)kDd @h  s *A? ?3"`? 2 0c{]"n`! 0c{]@ xmRKQōM BXhUo$nۥ6^^?!F%(=D&4μ"o{3{@l ^(BFFER7y)iFo&^b$ɈȐ@L55eP{,k)A3j3puS*պlAfQ< 5i͊!)ER;4$F AܤŃ3NV) ))Nu߮r^)염/~ i㾮t,M7:ͪߔHO+k)/9’] -s͵Ks?3mN&\Fsw=Hu[pPL ֚@߱ø}g2o†帳&Guju%1i]8ڸr5Dd h   s *A ? ?3"`? 2J{!90[XeL^&%n`!{!90[XeL^Z $@ȕ8xڵVMlW7oױA $VJc"$īT) T XUb'q1"Eɕz@zBz8 ġč7@4bfv86cw=~@J7 Q ]L(궠iƠrOhY%:%y0]89/)``aD߆,5,x>h|]jKIp`A;ۿgڏfIhVf3-8#hd'v do-65p[ \9HVnblŦc¸*O䙆zZJqr8IqrxuVV[-Dtvuv.p[]-ڰV嬉1dlZ/ cwRV ũu}un$>S@wl,‰H^Am6F P-<-zpԋ[™ŒcGv*N0hvDj2A>TvsavTTWqm6,Υ5LkWJu+є aMt&z9K{!~Law7į2aU&9ySvTG @yF(_⯼BP;Gʩ7dhzz NSS4͝<2<֮^ι 佞/">!ߗ DHzݑ|NyvPw$|,dҌ<޶ `l!iMw0C&I+`}J9(p?]5|qDZ_muĽ^yCo esO~U`3ҴM{}_ȷaHKXq .tLχFZz&N&x'+Y,J8OyvWz# $He=S'Ir@H$;S%'ayiPb4rxFEDFP/펀roA/'kG.* `!Y*tSQԖ~tԊa OB3:˰R-*ب;|az0/Jٕr*"Idr f/N `+Bu8&2b"ޫ#:̟ؤ!U,CBN%TԪWD,)(Y>#աTfb?BYs(egpمP-2,> &UxY&!EHX‡Bx{_\&6[Β\`o{Rxw(S:\-$R=~WA~C5c{['] kO,ht;Jc DQ6yIn< e2g]ٽUvxWяM,O+zZr"w@ho|61%ƃ3s(FQP,k3F7 /̧bNB'YHG-pZO[v[WkHovMzP$忀K CR䬌2B#,Q3sܓ t2i;KPxӇ+nąB% y%UzO!OVH֞Dq>`/۶^EZFM =@);Rd]rl9ќd2p8J&Z]rڢNA )rmtxQLB.ܸLi*z!KZv)ܖ"ajXOomR~owt#HA5}DyK _Ref527259914Dd KD  3 @@"?}DyK _Ref527259914Dd @h  s *A ? ?3"`?2 @ꢒҸb/68n`! @ꢒҸb/6 Ƚ xmRMK@}mj!UP?E1D h6ړ=xVуH< mTPLb@d4vA"NGy s.. Rd4ӣc Yf3af|taU9~XlbI>!ѐXC]WQ*rj9fQb,"9^",˜ܹlBkM nF^nR'|{W*=@d 2({{꾛w\ D$MBTOk[HDsŔuV9D.l3:1xX7Xt^zFA[evֶzX 6n6rң4LQΚ^SM~ݭȕlU_{bݮS}~~IDd @h  s *A ? ?3"`?2 @ꢒҸb/6;n`! @ꢒҸb/6 Ƚ xmRMK@}mj!UP?E1D h6ړ=xVуH< mTPLb@d4vA"NGy s.. Rd4ӣc Yf3af|taU9~XlbI>!ѐXC]WQ*rj9fQb,"9^",˜ܹlBkM nF^nR'|{W*=@d 2({{꾛w\ D$MBTOk[HDsŔuV9D.l3:1xX7Xt^zFA[evֶzX 6n6rң4LQΚ^SM~ݭȕlU_{bݮS}~~IDd yb  c $A ? ?3"`?23ӭf.$b- J>n`!ӭf.$b-   &xڭOkSAg%I$iUM JSA"b4XJmclȟD!/ԛ7[BOO!qviS%v2;Ap( ր7'uhJISˬy*k28O>ҴH^RЊ^s ʃԠ/ ~pYR?I5'ܜͤuw\O?| |3z^)).fuz p$yX0[ -FflWd\mCrV N "`Q&c1cTluZ9c8,Fם:t}6|qx4B0 Ox5z$iSDV[v9!~섴߲ޝ=RfǼUP8CT`5U᪱5Ъ!:g&SRXK浩} ~+DbN[F;&SbnI`$ħ Z&>eǽq0i1!M4F2FChb4.MɌgŽ*R2Q EȪ2xZ1ϵU ɦ@/#OLr*wFxYlrwyRԋ q5ԇ'I ?oy҆邖0XWBazpChm5{ མAV<0/Sw\۬}5&Dd b  c $A? ?3"`?2pŠ_/sm'L3Bn`!DŠ_/sm'@8 xڵUkA3M4%IkbiAт-$M=X)6c>FEZ)Dz /ē E<{*PHEƴٷ7og@ql T1CbZ k7tMsq}݊):J`jQ~D&|"\DY#j ^LY·/=wƲs.5޶j7fU |q0Z&iS$XI1f?|t+JzXf%h8-OhY=9Lb_yXGijʈ{gR [~Zfxmв=c ݕӹA|AmXV֯m ';G𜼚RC\r2Ůk9[Uj0s3Բdm4 gc]po`q_ƾw@JrRqU Uz"#=3B'2 U ] `.>uU@T(9>8wUh.FC0(vIxzwryų Ox>Њ.k)"ɛtDd b   c $A ? ?3"`?2-iOzinnq&m YFn`!iOzinnq&m`(PxڵUMhA~3VZ-6 (XжⵒF{Zl44iJRYŊT/("ţ9XB(E&nCZ)̼y|y3C 1 ?J%l6tZ:/|tFiP:>hrcj(RNHA/LXT,n>qJJ?h)D4I-BU=tO/ޏ%T yFFpÝQ#rt+:Dۻ0ԘH9/1Hf, C$F-Mx9Ѡ]`dҲF`7lv'"ОYe`p4]3jQ2˾;x OϤ7lC8-@ps0~-'gƕc 1՞ޏO H=lDYľ_nlj"y 떶bv;^S'w]dV4AWpE>Ae"J[* %cCI3࿚N~?r? fy=%=dT,ςBE_h=>G7nϒux4Njxl)eә̝SS˟qŘKfq-[Oh7*|N61_|b 1e,^hpCPzŠF|(+gJH>5X&߻( #EO:fȃq\X  qTT%"ʃ$$m<)lٴ3p[TuxOdw:G ᰼ɥ4O,mzSf?KͻS]nB]-.ZuA/)N}5aJ6Dd b   c $A ? ?3"`?2"8N8]֪k Nn`!8N8]֪kHhxڝRN@}NZR!p@ X@PDq`\RB$U)%|␿\@;vgyRvN@}HI P5&/#bD9F?C(Y +ZWe{Z!&lNHJQiqB9FG$QJIc,7t;]k53T{uV'<$FZ$7QQѿjD?K/[$&WƖ񎊭c✸dsgj0Ulm{n[&ʵum;5,f~an"`e32܆VboCe)f`% g:Hp3 /Dd).zHk]xLD ~} vy2P4_FypM{;-I\Dd b   c $A ? ?3"`?2"8N8]֪kPn`!8N8]֪kHhxڝRN@}NZR!p@ X@PDq`\RB$U)%|␿\@;vgyRvN@}HI P5&/#bD9F?C(Y +ZWe{Z!&lNHJQiqB9FG$QJIc,7t;]k53T{uV'<$FZ$7QQѿjD?K/[$&WƖ񎊭c✸dsgj0Ulm{n[&ʵum;5,f~an"`e32܆VboCe)f`% g:Hp3 /Dd).zHk]xLD ~} vy2P4_FypM{;-I\1TableSummaryInformation(jDocumentSummaryInformation8CompObjj՜.+,D՜.+,L px  MIT LCS%I  Introduction Title(_gwMTEquationNumberMTEquationSection(1.1)ti1.1V@V /#Normal$@`@a$ CJOJQJ_HaJmH sH tH ``` +H> Heading 1$$ & Fx@&a$5CJKH OJQJ\b`b +H> Heading 2$ & Fx@&5CJOJQJ\]^JaJB@B Heading 3$ & F@&5\R@R Heading 4$ & F<@&5CJ\aJT@T Heading 5 & F<@&56CJ\]aJN@N Heading 6 & F<@&5CJ\aJD@D Heading 7 & F<@&CJJ@J Heading 8 & F<@& 6CJ]T @T Heading 9 & F<@&CJOJQJ^JaJDA@D Default Paragraph FontViV  Table Normal :V 44 la (k@(No List LB@L Body Text 7$8$H$CJOJQJ^JaJe@ HTML Preformatted= 2( Px 4 #\'*.25@9dOJPJQJaJ@O@ MTEquationSection<B*^O"^ MTDisplayEquation !xx OJQJaJ.)@1. Page Number< @B< Footer  !CJaJF"@F Captiond`6CJ\aJV>@bV Title$<@&a$5CJ KHOJQJ\^JaJ HOrH References  & FP OJQJaJLC@L Body Text Indenth^h`6U@6 Hyperlink >*B*ph4@4 Header  !PR@P Body Text Indent 2 `.X@. Emphasis6]>@> j { Footnote TextCJaJ@&@@ Footnote ReferenceH*HH 5fh Balloon TextCJOJQJ^JaJDT@D 5fh Block Text x]^6P@6 5fh Body Text 2!x>Q@"> 5fh Body Text 3"xCJaJjM@2j 5fhBody Text First Indent#x7$8$H$` CJ^JaJXN@BX 5fhBody Text First Indent 2$x`TS@RT 5fhBody Text Indent 3%hx^hCJaJ2?@b2 5fhClosing &^<r< 5fh Comment Text'CJaJ@jqr@ 5fhComment Subject(5\$L@$ 5fhDate)RYR 5fh Document Map*-D M OJQJ^J<[@< 5fhE-mail Signature+<+< 5fh Endnote Text,CJaJh$@h 5fhEnvelope Address!-@ &+D/^@ OJQJ^JN%@N 5fhEnvelope Return.CJOJQJ^JaJ:`@: 5fh HTML Address/6]: : 5fhIndex 10^`: : 5fhIndex 21^`: : 5fhIndex 32^`: : 5fhIndex 43^`:: 5fhIndex 54^`:: 5fhIndex 65^`:: 5fhIndex 76^`:: 5fhIndex 87^`:: 5fhIndex 98p^p`H!H 5fh Index Heading95OJQJ\^J4/@4 5fhList:h^h`82@8 5fhList 2;^`83@8 5fhList 3<8^8`84@8 5fhList 4=^`85@8 5fhList 5>^`:0@: 5fh List Bullet ? & F>6@> 5fh List Bullet 2 @ & F>7@> 5fh List Bullet 3 A & F>8@"> 5fh List Bullet 4 B & F>9@2> 5fh List Bullet 5 C & FBD@BB 5fh List ContinueDhx^hFE@RF 5fhList Continue 2Ex^FF@bF 5fhList Continue 3F8x^8FG@rF 5fhList Continue 4Gx^FH@F 5fhList Continue 5Hx^:1@: 5fh List Number I & F>:@> 5fh List Number 2 J & F>;@> 5fh List Number 3 K & F><@> 5fh List Number 4 L & F >=@> 5fh List Number 5 M & F!- 5fh Macro Text6N$  ` @ @d`@a$OJQJ^J_HmH sH tH I@ 5fhMessage HeadergO8$d%d&d'd-DM NOPQ^8` OJQJ^J<^@< 5fh Normal (Web)POJQJ>@> 5fh Normal Indent Q^4O@4 5fh Note HeadingRDZ@2D 5fh Plain TextSCJOJQJ^JaJ0K@0 5fh SalutationT6@@R6 5fh Signature U^FJ@bF 5fhSubtitleV$<@&a$ OJQJ^Js 5fhTable 3D effects 1m:VWj#j#j#j#j.j.j.j. W$@d`@a$55\5B* \`J phB* `J phJJ 5fhTable 3D effects 2:VXj.@j#j9jj 4X$@d`@a$5\5\ 5fhTable 3D effects 3:VYj.@j j j#j9jj44Y$@d`@a$:B*`Jph B*`Jph5\5\vrv 5fhTable Classic 1:VZ0  j#j#j#jj Z$@d`@a$9B*`Jph6]5\56\]s 5fhTable Classic 2#:V[0  j% j#j0 jjj%  [$@d`@a$:5\B*`JphB* `J ph5\hth 5fhTable Classic 3:V\0    jj0  j0  \$@d`@a$QB* ph5B*\`JphB* `J ph56B*\]`Jphu 5fhTable Classic 4:V]0  jj0 j0 jj ]$@d`@a$X5\B* `J ph56B*\]`JphB* `J ph5\fvf 5fhTable Colorful 1:V^0    j% j% jj%  ^$@d`@a$<B*ph56\]56\]56\]hwh 5fhTable Colorful 2:V_0 j% jj0  j _$@d`@a$@56\]56B*\]`Jph56\]HxH 5fhTable Colorful 3:V`0j;$ j0 j%  `$@d`@a$5B*\`Jphy 5fhTable Columns 1 :Va0    j j jjjj#jj4a$@d`@a$l5\B*`Jph B*`Jph5\5\5\5\5\5\z# 5fhTable Columns 2:Vbj j jjjj% jj4b$@d`@a$5\B*`Jph B*`Jph5\5B*\`Jph5\B*`Jph5\5\{3 5fhTable Columns 3:Vc0j j jjj#j% j4c$@d`@a$h5\B*`Jph B*`Jph5\5\5\B*`Jph5\&|C& 5fhTable Columns 4:Vdj j jjj% 4d$@d`@a$LB*`Jph B*`Jph5\5\B*`Jpht}St 5fhTable Columns 5:Ve0    j jjj#j#4e$@d`@a$VB*`Jph B*`Jph5\5\5\56\]DcD 5fhTable Contemporary:Vf0j%@ j% j% 4f$@d`@a$<@B*`JphB*`Jph5B*\`Jphs 5fh Table Elegant_:Vg0j g$@d`@a$;B*`Jph~~ 5fh Table Grid7:Vh0h$@d`@a$~ 5fh Table Grid 1z:Vi0jj i$@d`@a$6]6]22 5fh Table Grid 2:Vj0jjj#j j$@d`@a$,5\5\5\5\ 5fh Table Grid 3:Vk0  jjj0  k$@d`@a$5\5\RR 5fh Table Grid 4:Vl0  jj0 j0  l$@d`@a$B5B*\`Jph5B*\`JphB*`Jph 5fh Table Grid 5:Vm0    jjj# j m$@d`@a$5\5\88 5fh Table Grid 6:Vn0    jj#j#j n$@d`@a$(5\B*`Jph5\^^ 5fh Table Grid 7:Vo0    jjj#j# j o$@d`@a$25\5\5\5\5\44 5fh Table Grid 8:Vp0jjj%  p$@d`@a$H5B*\`Jph5B*\`Jph5B*\`Jph 5fh Table List 1:Vq0  j%@ jj#j0 j4q$@d`@a$M@B*`JphB*`Jph56B* \]`J ph5\# 5fh Table List 2:Vr0 j%@ jj#j0 j4r$@d`@a$G@B*`JphB*`Jph5B*\`Jph5\&3& 5fh Table List 3:Vs0  j# j# j s$@d`@a$05B* \`J ph6B* ]`J phC 5fh Table List 4w:Vt0    j0   t$@d`@a$5B*\`JphS 5fh Table List 5:Vu0jj#  u$@d`@a$5\5\c 5fh Table List 6:Vv0j%@ j# j#  4v$@d`@a$5\5\s 5fh Table List 7:Vw0  j%@ j% jjj# j0  4w$@d`@a$>@B*`Jph5\5\5\5\ 5fh Table List 8:Vx0j%@ j% jjj#j0 4x$@d`@a$D@B*`Jph5\5\5\56\]T,T 5fhTable of Authoritiesy^`L#L 5fhTable of Figuresz ^`  5fhTable Professionall:V{0j%  {$@d`@a$5B*\`Jpho 5fhTable Simple 1:V|0  j#j# |$@d`@a$p 5fhTable Simple 2:V}j#j# j#j# j#j#}$@d`@a$O5\5\5B*\`Jph5\5\5\q 5fhTable Simple 3l:V~0    j%  ~$@d`@a$5B*\`Jph 5fhTable Subtle 1 :Vj0@ j# j# j0  j. jj4$@d`@a$5\5\ 5fhTable Subtle 2:V0j0  j0  j# j# jj$@d`@a$5\5\ 5fh Table Theme7:V0$@d`@a$# 5fh Table Web 1h:V03j $@d`@a$B*`Jph3 5fh Table Web 2h:V03j $@d`@a$B*`JphC 5fh Table Web 3h:V03j $@d`@a$B*`JphH.H 5fh TOA Headingx5OJQJ\^J&& 5fhTOC 1.. 5fhTOC 2 ^.. 5fhTOC 3 ^.. 5fhTOC 4 ^.. 5fhTOC 5 ^.. 5fhTOC 6 ^.. 5fhTOC 7 ^.. 5fhTOC 8 ^.. 5fhTOC 9 ^tOQt Style Caption + Left$ $xxa$56CJPJaJnHtH^O ^ !Style Style Caption + Left + BoldxO x Style Caption + Left Char.5CJOJPJQJ\_HaJmH nHsH tHdO ! d &Style Style Caption + Left + Bold CharOq2 CStyle References + Left: 0 pt Hanging: 12 pt After: 0 pt Line...( & F h$h^h`CJe-K'*.26789:-K '*.26789:= 6?J-K/x456o)=>?Hu c n S$n,QHp=-!!5"Q"u##$$&&''(@(_)a)b)c)d)e)f)))**+H+i+M,c,N../h/v22m3358879 :&:J;d;=@@mAxAABBCCDEEJFGG#HHHoI J JJJJJJJJJJJJJJJJJJJKK K K KKKKKKKKK"K#K&K'K(K)K*K+K.K0000000000000000 00u& 0u0u& 0u0u 00c 0c 0c 0c  0000000000000000      000000000000 00&0&0& 00(0(0(0(0(0(0(0(0(0(0(0(0(0(0( 00M,0M,0M,0M,0M,0M,0M,0M,0M,0M,0M,0M,0M,0M,0M, 00J;0J; 00@ 00@0@0@000@00@00@00@0H000H0H0A0@0p0eP@0@0@0@0@0@0@00ZP 00000000000000H00000000000000456o)=>?Hu c n S$n,QHp=-!!5"Q"u##$$&c,./h/v22m3358879 :&:d;@@mAxAABBCCDEEJFGG#HHHoI JJJJJ K#K.K 0000000000000000  0@ 02@ 020/  0@0 @0 @0 @ 0@0@0@0@0@0@0@0@0@0@0@0@0@0A 0@0i!@0i!@0i!@0i!@0i!@0i!@0i!@0i!@0_@0i!A 0A 0@0s@0s@0s@0_@0_@0s@0s@0s@0s@0s@0s@0s@0sO90L0O90L0O90L0Oy0L0Oy0L0O90L0O90L0O90L0@ 0 SO90L0O90L0O90L0@ 0i@ 0iOy00@ 0*iOy00@0O90P0O90P0O90P0@00"0@0 0XPO900 H`O900800!"0 @ vB: !+$&')+,/n1 36e79:ABHKOR-S*-/01245679:;<>?@ACDEFGIJL Q'07xIoQR+S-S+.38=BHKMN,S,5MOTln(@B , . !!!5"M"O"v###f$~$$$$$`&x&z&'''m)))I+a+c+....// /#/%/h///11121232v222m333888 :":$:f:~:::::-K:::::::::::::: :_ :::::::: !  ?2$,ˈ2"[T 2$Bxr ` 2$|rfX@H 0(  +@P   4(    ^ ! 3  s"*?`  c $X99?^ !   S  C"?   F L [/  [/ B2    [ s B2   l [] s B2   l ] -  B2   l ]  B2   >[0s B2   >0  B2   >0* B2   l=\S B2   l[\s B2   l\  B2   ?Z/r `B  c $D l  `B  c $D] > fB  s *D0l `B  c $Dtl  `B  c $DHl Z `B  c $D] > `B  c $D] >-  fB  s *D0- l-  fB  s *D0l `B  c $D\b `B  c $D\? `B  c $D\Hb    S C"?[    c $ C"? >    c $C"?w 7   S C"?] \(   S C"?   t  C C"?$ @   `B  c $D   6`B  c $D4 (;B2   "  (B2     B2   {-  S  C"? 1 +    S  C"? 1     S  C"?y;0  B S  ?_)./-Kc8t~*tt _Ref512889296 _Ref512874483 _Ref512868286 _Toc492242113 _Toc492245395 _Toc492242114 _Toc492245396 _Ref529169587 _Ref527259914c &&((f)/.Km ,&&?(?()&/.K/G\ 0GD1!1G!2G< 3G 4GtB!5G 6G^ 7GL 8Ga 9Gl:Gt#;G<"G|"?GN @GgAGegBG CG#DG/uEGlqFGd#GGD$uHG #BCCCFFGGGGGuHuHHII.K     "(( CCCCFFGGGHH~HHHII.K   9*urn:schemas-microsoft-com:office:smarttagsplace8*urn:schemas-microsoft-com:office:smarttagsCity=*urn:schemas-microsoft-com:office:smarttags PlaceName=*urn:schemas-microsoft-com:office:smarttags PlaceType9*urn:schemas-microsoft-com:office:smarttagsState>*urn:schemas-microsoft-com:office:smarttags PostalCode:*urn:schemas-microsoft-com:office:smarttagsStreet;*urn:schemas-microsoft-com:office:smarttagsaddressB *urn:schemas-microsoft-com:office:smarttagscountry-region   1456$$%%3388,:/:0:1:>>BB B)BBBBBBBBBBVCWCCCCgFoFG GGGHHJJJJ.K\!a!$$/ /&/'/_/d/?2E222BBBBBCgFrFJJJJ.K3333333333333)>?H8% n /,QHo-&&(@(f))M,c,/)/J;d;@@6AxArDDJJJJJJJJKK K K K KKKKKKKK!K#K%K.KJJJJ.KlmuilmuilmuilmuilmuilmuilmuilmuilmuiLIK MUI|nM}nL~2fKr{J^CR}BxAd@0CI(n ? Cax%p^)zRU~:,Y{K[Գ_P^F8jiraTrjXq !jot}?o1I[qҧMxvh$T}^`.^`.88^8`.^`. ^`OJQJo( ^`OJQJo( 88^8`OJQJo( ^`OJQJo(hh^h`. hh^h`OJQJo(^`o(()^`.pLp^p`L.@ @ ^@ `.^`.L^`L.^`.^`.PLP^P`L.h^`.h^`.hpLp^p`L.h@ @ ^@ `.h^`.hL^`L.h^`.h^`.hPLP^P`L.^`.^`.pLp^p`L.@ @ ^@ `.^`.L^`L.^`.^`.PLP^P`L.h^`OJQJo(hHh^`OJQJ^Jo(hHohpp^p`OJQJo(hHh@ @ ^@ `OJQJo(hHh^`OJQJ^Jo(hHoh^`OJQJo(hHh^`OJQJo(hHh^`OJQJ^Jo(hHohPP^P`OJQJo(hHh ^`OJQJo(h ^`OJQJo(oh pp^p`OJQJo(h @ @ ^@ `OJQJo(h ^`OJQJo(oh ^`OJQJo(h ^`OJQJo(h ^`OJQJo(oh PP^P`OJQJo(h ^`OJQJo(h ^`OJQJo(oh pp^p`OJQJo(h @ @ ^@ `OJQJo(h ^`OJQJo(oh ^`OJQJo(h ^`OJQJo(h ^`OJQJo(oh PP^P`OJQJo(h^`.h^`.hpLp^p`L.h@ @ ^@ `.h^`.hL^`L.h^`.h^`.hPLP^P`L.h ^`OJQJo(h ^`OJQJo(oh pp^p`OJQJo(h @ @ ^@ `OJQJo(h ^`OJQJo(oh ^`OJQJo(h ^`OJQJo(h ^`OJQJo(oh PP^P`OJQJo(^`OJPJQJ^Jo(- ^`OJQJo(o pp^p`OJQJo( @ @ ^@ `OJQJo( ^`OJQJo(o ^`OJQJo( ^`OJQJo( ^`OJQJo(o PP^P`OJQJo(P^`Po(.@@^@`o(.0^`0o(..``^``o(... ^`o( .... ^`o( ..... ^`o( ...... `^``o(....... 00^0`o(........hh^h`CJOJQJo([]^`o([]^`o(.pLp^p`L.@ @ ^@ `.^`.L^`L.^`.^`.PLP^P`L.hh^h`.h ^`OJQJo(h ^`OJQJo(oh pp^p`OJQJo(h @ @ ^@ `OJQJo(h ^`OJQJo(oh ^`OJQJo(h ^`OJQJo(h ^`OJQJo(oh PP^P`OJQJo(^`o(.^`.pLp^p`L.@ @ ^@ `.^`.L^`L.^`.^`.PLP^P`L.)rjI[qMxT}!jo!jo!jo!jo~:,_Prj8ji{KUCa?o!jop^rj!jo!jorj~}|!jo!jorjrj)rj!jo!joP                                                                               B34                J+        ?>#p1# } o}U3 t 4 P ss JuXxet?xe2I`F&(F|/, ='bo 8u!l$L%_(H),8*2C*OZ*%+}+D)-~-Xf/|/901P1H2wu2r35>5V5`56a6u{7AL8 :i:m:Q:2;7<+H>n=@A\A$mCODFjDlDE+E^`E`[F G~Hx/H05HJkuKKLQ:LJULfLMgPNRNxN)QiQ%IRZjSqSTSaT|cTRU(WqpXY~uY-Z\9\]x\f+]?4]^gd_!`"`_` aB7aJbbbwcZd%Idfh5fhl3ij j\kmknqo(op^pKrOrVr8Osu+vHwkw 9xT-ybyU