An Information Theory For Preferences

Size: px
Start display at page:

Download "An Information Theory For Preferences"

Transcription

1 An Information Theor For Preferences Ali E. Abbas Department of Management Science and Engineering, Stanford Universit, Stanford, Ca, Abstract. Recent literature in the last Maimum Entrop workshop introduced an analog between cumulative probabilit distributions and normalized utilit functions []. Based on this analog, a utilit densit function can de defined as the derivative of a normalized utilit function. A utilit densit function is non-negative and integrates to unit. These two properties of a utilit densit function form the basis of a correspondence between utilit and probabilit, which allows the application of man tools from one domain to the other. For eample, La Place s principle of insufficient reason translates to a principle of insufficient preference. The notion of uninformative priors translates to uninformative utilit functions about a decision maker s preferences. A natural application of this analog is a maimum entrop principle to assign maimum entrop utilit values. Maimum entrop utilit interprets man of the common utilit functions based on the preference information needed for their assignment, and helps assign utilit values based on partial preference information. This paper reviews maimum entrop utilit, provides aiomatic justification for its use, and introduces further results that stem from the dualit between probabilit and utilit, such as joint utilit densit functions, utilit inference, and the notion of mutual preference. INTRODUCTION In man decision situations, we are faced with multiple and conflicting attributes. When the decision situation is deterministic, each decision alternative is described b a single prospect (consequence). The problem of choosing the best decision alternative is that of ordering the prospects present or, alternativel, assigning a value function over the attributes of each prospect. The optimal decision alternative is the one that has the largest value as determined b the value function or the highest order in the ranked list. A normative justification for the order requirement is that a decision maker who cannot order the prospects is vulnerable to being a mone pump (we would choose prospect A over prospect B, but also prospect B over prospect A, and be willing to pa mone to move from one prospect to the other) When uncertaint is present, a decision alternative is characterized b several prospects and a probabilit distribution representing the probabilit of their occurrence. In this case, the rank order of the prospects alone is insufficient to determine the optimal decision alternative, and the Von Neumann and Morgenstern utilit values need to be elicited. [2]. To elicit these utilit values, a decision maker is first asked to order the prospects of the decision situation from best to worst. Once the order is provided, the net step is to assign a utilit value for each prospect. For an three ordered prospects, P P P 3, the decision maker assigns a probabilit, π, for 2 which she is indifferent to receiving for sure and a deal where she would receive P P2

2 with probabilit π and P with probabilit ( π ). If the utilit values of and P 3 P 3 are one and zero respectivel, then the utilit value of P 2, also called the preference probabilit of P 2, will be equal to π () + ( π)(0) = π. The higher the prospect is in the ranked list, the larger is the utilit value assigned to it. The optimal decision alternative is now the one, which has the highest epected utilit for its prospects. The normalization of the utilit values discussed above leads to an analog between utilit functions and cumulative probabilit distributions. Our goal in this paper is to review the analog between utilit and probabilit, and to draw on the applications of this analog to the maimum entrop principle and to man other concepts of information theor. The main requirement for this analog is that the decision maker can provide the complete preference order for the prospects of the decision situation she is facing. UTILITY ASSIGNMENT FOR ORDERED PROSPECTS We start this section with the definition of a utilit vector for a set of K ordered prospects of a decision situation. A utilit vector contains the utilit values of the prospects starting from lowest to highest. With no significant loss of generalit, we will assign a utilit value of zero to the least preferred prospect, u, and a value of one to the most preferred prospect, uk (we ignore the case of absolute indifference between the K prospects). The utilit vector has K elements defined as U ( u0, u, u2,..., uk 2, uk ) = (0, u, u2,..., uk 2,). () An utilit vector of dimension K can be represented as a point in a K-2 dimensional space in the region defined b 0 u u2... uk 3 uk 2. We will call this region the utilit volume. An eample of a two-dimensional utilit volume is shown in figure (). u 2 Utilit Volume 0 u u 2 0 Utilit Vector Representation FIGURE. A 2-dimensional utilit volume. u

3 The second definition is the utilit-increment vector, U, whose elements are equal to the difference between the consecutive elements in the utilit vector. The utilitincrement vector has ( K ) elements defined as U u u2 u uk 2 u u2 u3 uk ( 0,,..., ) = (,,,..., ). (2) Note that all elements of U are greater than or equal to zero and sum to one. Therefore, an utilit-increment vector can be represented as a point in a (K-) K dimensional simple : i =, i 0 i=. These two properties of the utilit increment vector over the simple form the basis of the analog between probabilit and utilit and will be used further in this paper. We will refer to this simple as the utilit simple. For eample, a 3-dimensional utilit simple is shown in figure (2). u 3 u u 2 FIGURE 2. A 3-dimensional utilit simple. Note that all points in the utilit volume or the utilit simple satisf the decision maker s preference ordering of the prospects but assign different utilit values to them. In other words, knowledge of the preference order alone tells us nothing about the location of the utilit vector over the utilit volume or the utilit increment vector over the utilit simple. If all we know about the decision maker s preferences is the ordering of the prospects, it is reasonable to assume that the location of the utilit increment vector is uniforml distributed over the utilit simple. This is also the maimum entrop distribution over the simple. The uniform distribution over the utilit simple implies that the location of the utilit increment vector is determined b a Dirichlet distribution whose K- parameters are equal to one. This is in turn implies the following three propositions described in []. Proposition : Marginal Distributions of the Utilit Vector The maimum entrop marginal distributions for the utilit values of a set of K ordered prospects are the famil of Beta distributions, Beta( j, K j ), j =,... K 2.

4 Given the marginal distribution for each element in the utilit vector, it is known that the logical assignment of each utilit increment value is the mean of its marginal j distribution [3]. The mean of a Beta( j, K j ) distribution is equal to, and is K the utilit value ou would assign to each prospect, j. The utilit values for prospects j=0 and j=k- are deterministic with values 0 and respectivel. Proposition 2: Marginal Distributions of the Utilit Increment Vector The maimum entrop marginal distributions for the elements of the utilit increment vector have identical maimum entrop marginal distributions, Beta (, K - 2). Given the marginal distribution for each element in the utilit increment vector, it is known that the logical assignment of each utilit increment value is the mean of its marginal distribution [3]. The mean of a Beta (, K - 2) distribution is K. We now have a maimum entrop assignment for the location of the utilit increment vector and are read for proposition 3 below. Proposition 3: Utilit Increment Vector Assignment The maimum entrop assignment for increments in the utilit values of a set of K ordered prospects ields identical increments in utilit values equal to K. For eample, if we have a decision situation with five ordered prospects, A B C D E, the maimum entrop assignment of the utilit increment vector is U = (0.25, 0.25, 0.25, 0.25) and the maimum entrop assignment of the utilit vector is U = (0,0.25, 0.5, 0.75,). Note that using the definitions of utilit volume and utilit simple, as well as the maimum entrop uniform joint distribution, we have made an unbiased assignment of utilit values for five prospects when onl the preference order is known. Note also that this unbiased assignment gave equal increments in utilit values. Now we are read for the first application of our utilit-probabilit analog: to assign unbiased utilit values when onl the order of the prospects is available. We mean b unbiased utilit values, those that do not lead to arbitrar assumptions of preference information that is not available. THE PRINCIPLE OF INSUFFICIENT PREFERENCE In the previous section we illustrated how to assign utilit values for K prospects of a decision situation when onl the preference order is available. Now we show how to

5 work through the same problem using our utilit-probabilit analog. The analogous problem in probabilit land is to assign a probabilit to the outcomes of a discrete random variable when no further information is available. This problem dates back to Laplace s principle of insufficient reason. Laplace suggested that we assign equal probabilities to all outcomes unless there is information that suggests otherwise. The utilit-probabilit analog then suggests a principle of insufficient preference, which can be stated as follows: when onl the preference order of the prospects is known, we assign equal increments in utilit values unless there is preference information that suggests otherwise. Note that the principle of insufficient preference provides the same assignment of utilit values as suggested b proposition 3 above. The principle of insufficient preference assigns utilit values for discrete ordered prospects. In man cases, however, we would like to assign utilit values over a continuous domain. In the net section we will etend our analsis to the continuous case and present an additional element to our utilit probabilit analog: a utilit densit function. UTILITY FUNCTIONS AND UTILITY DENSITY FUNCTIONS In the continuous case the utilit vector has an infinite number of elements tracing a utilit curve over the domain of monetar prospects. The largest value is still one and the lowest value is zero corresponding to the utilit values of the best and least preferred prospects respectivel. A utilit curve thus has the same mathematical properties as a cumulative distribution function. The utilit-increment vector is now the derivative of the utilit curve and represents a utilit densit function, u ( ), which is analogous to a probabilit densit function in that it is non-negative and integrates to unit. U( ) u ( ) (3) Our problem of assigning utilit values for continuous prospects of a decision situation, when partial preference information is available, is now equivalent to that of assigning an unbiased utilit densit function for the ordered prospects. The latter problem is analogous to the problem of assigning prior probabilities to a continuous random variable when onl partial information is available. In this paper we will generalize the utilit assignment problem and present a maimum entrop approach for the assignment of utilit values when onl partial preference information is available. First let us discuss the entrop of a utilit function and its interpretation.

6 THE ENTROPY OF A UTILITY FUNCTION Having defined the notion of a utilit densit function, it is natural to etend our analsis to the entrop measure applied to a utilit function, and discuss its interpretations. The differential entrop term applied to a utilit densit function is h ( ) = u ( )ln( u ( )) d (4) Let us take, as an eample, an eponential utilit densit function over a bounded domain [a, b]. We have γ γ e u ( ) =, a b γa γb e e (5) where γ is the decision maker s risk aversion coefficient. As γ 0, the decision maker becomes risk neutral (he values uncertain deals at their epected value). Using L hopital s formula, we can also show that γ γ e γa γb e e b a as γ 0. (6) The utilit densit function approaches a uniform densit and the entrop of the utilit function thus achieves its maimum value of ln( b a). Note that the uniform utilit densit function matches the principle of insufficient preference discussed earlier as it provides equal increments in utilit values for equall spaced prospects. The utilit function that corresponds to this uniform utilit densit function can be obtained b integration and is linear. The linear utilit function is thus the maimum entrop utilit function over a continuous bounded domain. As γ +, γ γ e δ ( a) γa γb e e (7) The utilit densit function converges to an impulse utilit densit function and the entrop approaches a minimum value of negative infinit showing less spread in the utilit densit. The impulse utilit densit function is the minimum entrop utilit function and integrates to a step utilit function. As γ, γ γ e δ ( b) γa γb e e (8)

7 The utilit densit function converges again to an impulse utilit densit function, and the entrop approaches a minimum value of negative infinit. This is the other etreme where the step in the utilit function occurs at the upper bound. The step utilit function is thus the minimum entrop utilit function and corresponds, in this eample, to cases of etreme risk aversion or risk seeking behavior. Figure 3 shows an eample of two eponential utilit densit functions for different values of γ and different entrop values. E ponential Utilit Densit Function Eponential Utilit Densit Function Variable Variable FIGURE 3. (a) Utilit densit function for small values of γ is approimatel uniform showing large spread. (b) Utilit densit function for large values of γ converges to an impulse utilit densit showing less spread and integrates to a step utilit function. A step utilit function appears often in behavioral decision making literature and is known as the aspiration utilit function. Simon [4] suggested that individuals could simplif decision problems b having binar goals: satisfactor if the outcome is above the aspiration level or unsatisfactor if it is below it. Aspiration Utilit Function Unsati sfactor Satisfactor Aspiration Level Million Dollars ($) FIGURE 4. Aspiration (Minimum Entrop) Utilit Function

8 The interpretation of entrop for utilit functions is thus a measure of spread about a decision maker s aspiration level. Aspiration utilit functions have the least entrop in their utilit densit functions showing binar preferences, while uniform utilit densit functions have the highest entrop showing smoother utilit functions that var graduall over a bounded domain. MAXIMUM ENTROPY UTILITY FUNCTIONS Now we will make use of the concept of the entrop of a utilit function and present a method to assign unbiased utilit values based on the partial preference information we know about the decision maker. B partial preference information, we mean an information that includes the complete preference order of the prospects but does not include knowledge of the complete utilit values. In the probabilit domain, Edwin Janes [5] proposed the maimum entrop principle suggesting the use of a probabilit distribution, which has maimum entrop subject to whatever information is known. In an analogous manner, we defined the maimum entrop utilit principle, [], that suggests the use of the utilit function (or utilit vector) whose utilit densit function (or utilit increment vector) has maimum entrop subject to whatever preferences are known. This result is an etension of the principle of insufficient preference described earlier. Maimum entrop utilit provides an interpretation for man of the utilit functions used in practice based on the preference information constraints needed for their assignment. For eample when onl bounds on the domain of continuous outcomes are available, the maimum entrop utilit densit function is uniform and integrates to a risk neutral utilit function as we have discussed earlier. When the first moment of the utilit function is known over a positive domain, the maimum entrop utilit function is eponential with the first moment equal to the risk tolerance. This provides us with a new interpretation for the first moment of an eponential utilit function over a positive domain. When the first and second moments are known, the maimum entrop utilit function over a continuous domain leads to the famous S- shaped prospect theor utilit function [6]. Maimum entrop utilit presents a method to assign unbiased utilit values when partial preference information is available. For eample, the maimum entrop utilit function given knowledge of some utilit values leads to a piecewise linear utilit function that passes through the given utilit points. UNINFORMATIVE UTILITY FUNCTIONS Much of the research on uninformative and entropic priors for probabilit functions can be translated through our utilit-probabilit analog into similar research on uninformative utilit functions. For eample, when Jeffres [7] uninformative prior, u ( ) =, > 0is used as an uninformative utilit densit function, it leads to the

9 famous logarithmic utilit function that is used ver often to represent the decision maker s preferences in the literature. ATTRIBUTE DOMINANCE UTILITY FUNCTIONS In this section we present the etension of the previous analsis to multiattribute utilit functions with ordered attributes. We assume that ( min, ) min is the least preferred prospect and (, ) is most preferred. Furthermore we will assume that ma ma the values of (, ) are arranged such that for an 0,, 0, we have > (, ) (, ) and > (, ) (, ) (9) With no loss of generalit, we will use a normalized multiattribute utilit function, U (, ), over the attributes in all of our analsis. i.e. Based on this formulation, we have 0 U (, ) (0) U (, ) = 0, U (, ) = () min min ma ma Now we will focus on a class of multiattribute utilit functions where a prospect (, ) is a least preferred prospect if either or is at its minimum value. This requirement places the following constraints on the utilit function: U (, ) = U (, ) = U (, ) = 0 [, ], [, ] min min min min min ma min ma (2) We will call the multiattribute utilit functions that satisf condition (2) attribute dominance utilit functions since an attribute set at a minimum dominates the remaining attributes and sets the multiattribute utilit function to a minimum. For attribute dominance utilit functions we will define the marginal utilit function over a single attribute,, as the numerical value of the multiattribute utilit function when all other attributes are set at their maimum values. I.e. for two attributes we have U ( ) U (, ) ma (3) Now we define a conditional utilit function, U ( ), as U U (, ) U (, ) ( ) = (4) U ( ) U (, ) ma

10 A conditional utilit function, U ( ), is the utilit function for attribute when we are guaranteed a fied amount of attribute. The conditional utilit function U ( ) is the path traced on the multiattribute utilit function for different values of and a fied value of. Equation (4) shows that for attribute dominance utilit functions, the minimum value of the conditional utilit function, U (, is min ) equal to zero and the maimum value, U ( ma ), is equal to one. The denominator in equation (4) thus serves as a normalizing term so a conditional utilit function can be assessed directl from the decision maker on a scale of 0 to. Re-arranging equation (4) gives: U (, ) = U ( ) U ( ) (5) In a similar manner, we can also define U ( ) as U U (, ) ( ) U ( ) (6) Combining equations (4) and (6), we have U ( ) U ( ) U ( ) = (7) U ( ) Equation (7) provides a method to update our utilit values over one attribute when we are guaranteed a fied amount of another. We will call equation (7) Baes rule for utilit inference. Now define a joint utilit densit function, [, ] and [, ] as: min ma min ma u (, ), over two attributes, u 2 U(, ) (, ) (8) where U (, ) is the multiattribute utilit function for the two attributes and. From the properties of derivatives, the integral of a joint utilit densit function that is defined b equation (8) over a region a b, c d is d b c a u (, ) dd = U ( b, d) U ( a, d) U ( b, c) + U ( a, c) (9) If we set a = min and c= min, and if attribute dominance utilit eists, then equation (9) reduces to

11 U (, ) u (, ) dd = min min (20) Equation (20) illustrates how attribute dominance utilit functions can be obtained b integrating a joint utilit densit function from the least preferred prospect to a prospect (, ). This integral provides an analog with joint cumulative distribution functions obtained b integrating a joint probabilit densit function over the domain of the variables. We note that an general multiattribute utilit function can be decomposed into a linear combination of attribute dominance utilit functions. For eample, rearranging equation (9) and using an general prospect b=, d=, a = min and c= min gives U (, ) U (, ) U (, ) U (, ) u (, ) dd = + + (2) min min min min min min The terms U (, ) and U (, are scaled attribute dominance utilit min functions of one dimension since the utilit function is zero if that single attribute is set to a minimum, and the scaling constants are equal to U (, and U (, min ma ) respectivel. Now we discuss the integral term min ) I(, ) u ( dd, ) = min min ma min ). We note that, b properties of a definite integral, I( min, ) = I(, min ) = 0. In other words, this integral has a value of zero if either or is at the minimum value. Furthermore, this integral has a maimum value equal to k = I( ma, ma ). Let us normalize this integral term and define C( as, ) C (, ) min min ma ma min min u (, ) dd I(, ) = I( ma, u (, ) dd From equation (22), we note that C( has, ) the same mathematical properties as an attribute dominance utilit function. In other words we can write ma min min ) (22) U (, ) = kc (, ) + kc (, ) + k C (, ) (23) where are scaling constants and C(, ), C (, ), and C (, ) are k, k, and k min min normalized attribute dominance utilit functions.

12 Similarl, for three dimensions we can define and therefore, u z 3 U(,, z) (,, z), (24) z U (,, z) = U (,, z ) + U (,, z ) + U (,, z) z z min min z min min z min min U (,, z) U (,, z) U (,, z ) z z min z min z min + zmin min min u z (,, z) dddz (25) Once again we note that the terms Uz (, min, zmin ), Uz ( min,, zmin ), and U (,, z) are scaled attribute dominance utilit functions of one dimension, z min min C(,, z ) = z zmin min min u z (,, z) dddz is a scaled attribute dominance utilit function of three dimensions, and that Uz ( min,, z), Uz (, min, z), and Uz (,, zmin ) can be reduced to one dimensional and two dimensional attribute dominance utilit functions using the arguments for the two dimensional case presented above. In other words, we can write equation (25) as U ( z,, ) = kc (,, z ) + kc (, z, ) + kc (,, z) z min min min min z min min k C(,, z) k C(,, z) k C(,, z ) + k z min z min min z C(,, z) (26) Equation (26) epresses an general three-dimensional multiattribute utilit function in terms of a linear combination of attribute dominance utilit functions. B induction, we can show that the same formulation applies to higher dimensions. We note that a special case of decomposing utilit functions into attribute dominance utilit functions arises when the attributes have mutual utilit independence and leads to the multilinear utilit function. In order to see this, we recall the following theorem from about mutual utilit independence [8]. If two attributes, and, have mutual utilit independence, their multiattribute utilit function has the multilinear form U (, ) = k U ( ) + k U ( ) + k U ( ) U ( ) (27) where k, k, and k = k k are constants, and U ( ) and ( ) are the utilit U functions of attributes and respectivel. If we substitute for C (, ) = UU ( ) ( ) into (23) we note that (27), is indeed a special case.

13 MAXIMUM ENTROPY ATTRIBUTE DOMINANCE UTILITY FUNCTIONS B analog with joint probabilit densit functions, we can now we etend the maimum entrop utilit analsis using the joint utilit densit function that we have developed. For eample, if the decision maker can provide onl the marginal utilit function for each attribute present, then the maimum entrop formulation maimizes the entrop of the joint utilit densit function subject to the given preference information constraints. If for eample, we have two attributes and that are defined on the unit square, and we know the marginal utilit functions, the formulation for the problem is ma u(, )ln( u(, )) dd st 0 udd (, ) = U( ), udd (, ) = U( ), udd (, ) =. (28) The solution to this problem provides a joint utilit densit function that is equal to the product of the marginal utilit densit functions. u (, ) = u( u ) ( ), 0,0. (29) The joint utilit densit function integrates, using equation (20), to U(, ) = U ( ) U ( ), 0,0. (30) The product form of the utilit densit functions presented in equation (29) is the condition of utilit independence that is analogous to probabilit independence and applies to an general class of multiattribute utilit functions that have mutual utilit independence. We can show this b substituting equation (27) into equation (8) to get 2 U(, ) u (, ) = ku( u ) ( ) (3) If the marginal utilit functions are known, the maimum entrop attribute dominance utilit function assumes utilit independence between the attributes. If

14 more information is available about the utilit function, such as moments, risk aversion coefficients, or utilit dependence parameters, it can also be incorporated into the maimum entrop formulation and an attribute dominance utilit function can be obtained b integration. We recall that the construction of an general multiattribute utilit function involves the construction of several attribute dominance utilit functions of different dimensions. RELATIVE ENTROPY The KL-distance (often known as the relative entrop or cross-entrop) is a measure of divergence between two probabilit distributions. In a similar manner, the KL-distance is a measure of the divergence between two utilit densit functions. The epression for the KL-distance between a utilit densit functions, u() and another utilit densit function q() is given as u ( ) KL distance = u( )ln( ) (32) q ( ) The maimum entrop utilit principle can now be generalized b analog to a minimum cross-entrop utilit principle b minimizing the KL-distance relative to a known target utilit densit function. This helps incorporate information about the shape of the utilit function or its relation to a famil of utilit functions into the maimum entrop formulation. UTILITY DEPENDENCE: AN INFORMATION THEORETIC POINT OF VIEW In the previous sections we discussed the interpretation of the entrop and relative entrop as the appl to utilit functions. In this section we discuss the interpretation of appling several measures of dependence that were developed in information theor into measurements of utilit dependence. We provide some insights on their dual interpretations and an intuitive meaning for utilit dependence that results from the analog with joint probabilit functions. We start with the notion of conditional entrop. Conditional Entrop Having discussed a joint utilit densit function, and a conditional utilit function, we can now define the conditional entrop of attribute Y given a guaranteed amount of attribute X as hy ( X) = u(, )ln( u( )), (33)

15 where u ( ) = U ( ) is the conditional utilit densit function of for a given value of. It is known from information theor that the conditional entrop is less than or equal to the marginal entrop, with equalit if and onl if the two variables are independent. I.e. hy ( X) hy ( ). (34) The interpretation for this result when applied to utilit theor is that our utilit functions for attributes have less entrop (become closer to aspiration utilit functions) when conditioned on another attribute. However, when the two attributes are utilit independent, the conditional entrop remains the same as the marginal entrop, since the utilit functions for each attribute are not updated. Mutual Preference The notion of mutual information translates using the utilit-probabilit analog into mutual preference. In probabilit land, the mutual information between two variables is the KL-distance between their joint distribution and the product of their marginal distributions. The interpretation for the mutual information is the amount of information induced about a random variable, X, when we know the outcome of another random variable, Y. Let us now consider the two etreme cases of probabilistic dependence. The first case is when the two variables are independent. In this case we have a marginal probabilit distribution for each variable. Knowing the outcome of one variable, X, tells us nothing about the other variable Y, and so this knowledge does not update its probabilit distribution. Therefore the entrop of Y remains unchanged whether or not we know X. The second case is when the two variables are deterministicall related through a functional epression. For eample, Profit = 0.5*Revenue. In this case, knowledge of the Revenue tells us the eact value of profit. If we know revenue, the distribution of profit is updated to a step distribution function (or an impulse probabilit densit function) at that particular value of profit. The entrop of the updated distribution of X is reduced to negative infinit (it is now deterministic) once we know the outcome of Y. The mutual information between the two variables in this case is infinite as knowledge of one variable provides infinite reduction in the entrop of the other. Now we consider the analogous interpretation of mutual information when applied to utilit, which we call mutual preference. If we have a prospect with two attributes, we have a utilit function for each attribute. Now we consider two etreme cases. The first is when the attributes have utilit independence. If we are now guaranteed a fied amount of one attribute, it does not change our utilit function over the other (b definition of utilit independence), and therefore this guarantee does not change its entrop. The mutual preference between the attributes is therefore equal to zero. On the other hand, the attributes ma be deterministicall related. Once again, we

16 consider the eample Profit = 0.5*Revenue. In this case, if we are guaranteed a fied amount of Revenue, sa R 0, then we automaticall set a reference point for our satisfaction with profit, 0.5* R 0. An value of profit that is below 0.5* R 0 will be unsatisfactor for us while an value of profit that is above 0.5* R 0 will be satisfactor. In other words, we now have a step (aspiration) utilit function (or an impulse utilit densit function) for profit. A guarantee of the Revenue attribute has sharpened our utilit function for profit and it is now a target or aspiration level. The mutual preference between the two attributes is thus infinite since guarantee of one attribute reduces the entrop in the utilit function of the other to its minimum value of negative infinit. AXIOMATIC JUSTIFICATION OF THE MAXIMUM ENTROPY UTILITY PRINCIPLE We have discussed the maimum entrop utilit principle and the application of several measures of information theor to utilit theor. In this section we further draw on this analog to provide an aiomatic derivation for the maimum entrop utilit principle. Shore and Johnson [9] showed that Janes maimum entrop principle for probabilit inference satisfies a set of reasonable aioms that stem from one fundamental principle: if a problem can be solved in more than one wa, the results should be consistent. The also showed that maimizing an function but the entrop will lead to inconsistencies in these aioms unless that function and the entrop have identical maima (an monotonic function of the entrop will work). The analog between probabilit and utilit densit functions that we have developed allows the application of their analsis directl to the maimum entrop utilit principle. Now we will mention each of the aioms and discuss how the relate to utilit assignment. Aiom : Uniqueness The utilit values that we assign b maimum entrop utilit should be unique. The justification for this is that if we do not have unique utilit values, we are vulnerable to being a mone pump (as discussed earlier). Another justification for the uniqueness of the maimum entrop utilit assignment is that if we solve the utilit assignment problem twice in eactl the same wa, we epect the same utilit values to be assigned in both times. Aiom 2: Invariance We should epect the same utilit values from the maimum entrop utilit principle when we solve the problem in two different coordinate sstems. For eample, if a utilit densit function is assigned in Cartesian co-ordinates and

17 integrate to get a multiattribute utilit function, we should get the same results if we solve the problem in polar coordinates and integrate over the domain of the polar coordinates to get a utilit function. Aiom 3: Sstem Independence Consider two independent utilit functions or two attributes that have utilit independence. If we receive information about the two attributes, then we would like the joint utilit densit function that is assigned to be equal to the product of the individual utilit densit functions (impling utilit independence as discussed above). Aiom 4: Subset Independence This aiom concerns itself with situations where we receive preference information about the conditional utilit functions in attribute dominance utilit. One wa of accounting for this new information is to update the conditional utilit functions U(B A) individuall. Another wa is to marginalize the attribute dominance utilit function to get U(B) and then appl the new preference information to U(B). The results should be consistent in both cases. As mentioned above, if these aioms are required b a utilit assignment mechanism, then maimizing an function but the entrop will lead to inconsistencies in these aioms unless that function and the entrop have identical maima. In addition to satisfing the previous aioms, the maimum entrop utilit principle satisfies two essential desiderata. The first desideratum, as mentioned earlier, is that for different people with the same order of preferences for the prospects and where we have the same preference information we should assign the same utilit values. The second desideratum is -utilit and probabilit independence- that results from the foundations of normative utilit theor. The utilit value of a prospect should not depend on the probabilit of getting that prospect due to the normative separation of beliefs from preferences [0]. CONCLUSIONS In this paper we have provided an analog between probabilit and utilit and built on the notion of a utilit densit function to a joint utilit densit function for multiple attributes. We applied concepts of information theor to utilit theor and gained insights on the notions of utilit dependence and aspiration utilit functions. The analog between probabilit and utilit provides several directions for future research. For eample, analogous interpretations for Co s aioms provide the basis for an aiomatic derivation of utilit inference. The algebra of probable inference translates to the algebra of preferences for logical operations on the attributes. The idea of Influence diagrams as graphical representations of joint probabilit distributions [] etends to the notion of utilit diagrams, which are graphical representations of multiattribute utilit functions and epress the dependence relations between the attributes. The method of copulas for constructing joint cumulative

18 distribution functions translates into a method for constructing attribute dominance utilit where the dependence parameters can be determined b the trade-off coefficients between the attributes. Assessing utilit moments and utilit crosscorrelation coefficients can also be used to construct a multiattribute utilit function. Interchanging utilit functions with probabilit functions provides a wealth of new results and insights. For eample, the notion of utilit dominance is the analog of stochastic dominance [2]. ACKNOWLEDGEMENTS I would like to thank Ronald Howard and Mron Tribus for introducing me to the Maimum Entrop communit and to the work of Edwin Janes and Richard Co. REFERENCES. A.E. Abbas. An Entrop Approach for Utilit Assignment in Decision Analsis. In Baesian Inference and Maimum Entrop Methods in Science and Engineering, Moscow, ID. AIP Conference Proceedings, C. Williams (ed.) vol 659. (2002). 2. J. Von Neumann, and O. Morgenstern. Theor of games and economic behavior. Princeton, N.J. Princeton Universit. 2 nd edition. (947). 3. R.A. Howard. Decision Analsis: Perspectives on Inference, Decision, and Eperimentation. Proceedings of the IEEE. Vol 58. No.5 pp (970). 4. Simon, H.A. A behavioral model of rational choice. The Quarterl Journal of Economics 69, (955). 5. E. Janes. Information theor and statistical mechanics. Phsical Review, vol 06, pp , (957). 6. D. Kahneman and A. Tversk. Prospect Theor: An analsis of decision under risk. Econometrica. 47, (979). 7. H. Jeffres. Theor of Probabilit. Oford Universit Press, London, (96). 8. R. Keene and H. Raiffa. Decisions with Multiple Objectives: Preferences and Value Tradeoffs. John Wile and Sons, Inc. (976). 9. J. Shore and R. Johnson. Aiomatic Derivation of the Principle of Maimum Entrop and the Principle of Minimum Cross-Entrop. IEEE Transactions on Information Theor. Vol 26, No., pp (980). 0. P. Samuelson. Probabilit, Utilit, and the Independence Aiom. Econometrica.20, (952).. R.A. Howard and J.E. Matheson. Influence Diagrams, in The principles and Applications of Decision Analsis, Vol. II, R. A. Howard and J. E. Matheson (eds.). Strategic Decisions Group, Menlo Park, California, USA. (984). 2. A. E. Abbas. Entrop Methods in Decision Analsis. PhD dissertation. Stanford Universit. (2003). 3. A. E. Abbas. Maimum Entrop Utilit. Submitted to Operations Research. (2002).

Ordinal One-Switch Utility Functions

Ordinal One-Switch Utility Functions Ordinal One-Switch Utilit Functions Ali E. Abbas Universit of Southern California, Los Angeles, California 90089, aliabbas@usc.edu David E. Bell Harvard Business School, Boston, Massachusetts 0163, dbell@hbs.edu

More information

2.3 Quadratic Functions

2.3 Quadratic Functions 88 Linear and Quadratic Functions. Quadratic Functions You ma recall studing quadratic equations in Intermediate Algebra. In this section, we review those equations in the contet of our net famil of functions:

More information

You don't have to be a mathematician to have a feel for numbers. John Forbes Nash, Jr.

You don't have to be a mathematician to have a feel for numbers. John Forbes Nash, Jr. Course Title: Real Analsis Course Code: MTH3 Course instructor: Dr. Atiq ur Rehman Class: MSc-II Course URL: www.mathcit.org/atiq/fa5-mth3 You don't have to be a mathematician to have a feel for numbers.

More information

Practice Problems **Note this list of problems is by no means complete and to focus solely on these problems would be unwise.**

Practice Problems **Note this list of problems is by no means complete and to focus solely on these problems would be unwise.** Topics for the Final Eam MATC 100 You will be allowed to use our MATC 100 calculator. The final eam is cumulative (Sections.-., Sections 3.1-3.5, Sections.1-.5) - see the details below. Sections.-. & 3.1-3.3:

More information

1.6 CONTINUITY OF TRIGONOMETRIC, EXPONENTIAL, AND INVERSE FUNCTIONS

1.6 CONTINUITY OF TRIGONOMETRIC, EXPONENTIAL, AND INVERSE FUNCTIONS .6 Continuit of Trigonometric, Eponential, and Inverse Functions.6 CONTINUITY OF TRIGONOMETRIC, EXPONENTIAL, AND INVERSE FUNCTIONS In this section we will discuss the continuit properties of trigonometric

More information

8.4. If we let x denote the number of gallons pumped, then the price y in dollars can $ $1.70 $ $1.70 $ $1.70 $ $1.

8.4. If we let x denote the number of gallons pumped, then the price y in dollars can $ $1.70 $ $1.70 $ $1.70 $ $1. 8.4 An Introduction to Functions: Linear Functions, Applications, and Models We often describe one quantit in terms of another; for eample, the growth of a plant is related to the amount of light it receives,

More information

Chapter Nine Chapter Nine

Chapter Nine Chapter Nine Chapter Nine Chapter Nine 6 CHAPTER NINE ConcepTests for Section 9.. Table 9. shows values of f(, ). Does f appear to be an increasing or decreasing function of? Of? Table 9. 0 0 0 7 7 68 60 0 80 77 73

More information

LESSON #11 - FORMS OF A LINE COMMON CORE ALGEBRA II

LESSON #11 - FORMS OF A LINE COMMON CORE ALGEBRA II LESSON # - FORMS OF A LINE COMMON CORE ALGEBRA II Linear functions come in a variet of forms. The two shown below have been introduced in Common Core Algebra I and Common Core Geometr. TWO COMMON FORMS

More information

8. BOOLEAN ALGEBRAS x x

8. BOOLEAN ALGEBRAS x x 8. BOOLEAN ALGEBRAS 8.1. Definition of a Boolean Algebra There are man sstems of interest to computing scientists that have a common underling structure. It makes sense to describe such a mathematical

More information

Information, Utility & Bounded Rationality

Information, Utility & Bounded Rationality Information, Utility & Bounded Rationality Pedro A. Ortega and Daniel A. Braun Department of Engineering, University of Cambridge Trumpington Street, Cambridge, CB2 PZ, UK {dab54,pao32}@cam.ac.uk Abstract.

More information

Biostatistics in Research Practice - Regression I

Biostatistics in Research Practice - Regression I Biostatistics in Research Practice - Regression I Simon Crouch 30th Januar 2007 In scientific studies, we often wish to model the relationships between observed variables over a sample of different subjects.

More information

CHAPTER 2: Partial Derivatives. 2.2 Increments and Differential

CHAPTER 2: Partial Derivatives. 2.2 Increments and Differential CHAPTER : Partial Derivatives.1 Definition of a Partial Derivative. Increments and Differential.3 Chain Rules.4 Local Etrema.5 Absolute Etrema 1 Chapter : Partial Derivatives.1 Definition of a Partial

More information

6.4 graphs OF logarithmic FUnCTIOnS

6.4 graphs OF logarithmic FUnCTIOnS SECTION 6. graphs of logarithmic functions 9 9 learning ObjeCTIveS In this section, ou will: Identif the domain of a logarithmic function. Graph logarithmic functions. 6. graphs OF logarithmic FUnCTIOnS

More information

3.0 PROBABILITY, RANDOM VARIABLES AND RANDOM PROCESSES

3.0 PROBABILITY, RANDOM VARIABLES AND RANDOM PROCESSES 3.0 PROBABILITY, RANDOM VARIABLES AND RANDOM PROCESSES 3.1 Introduction In this chapter we will review the concepts of probabilit, rom variables rom processes. We begin b reviewing some of the definitions

More information

1.1 The Equations of Motion

1.1 The Equations of Motion 1.1 The Equations of Motion In Book I, balance of forces and moments acting on an component was enforced in order to ensure that the component was in equilibrium. Here, allowance is made for stresses which

More information

Math 20 Spring 2005 Final Exam Practice Problems (Set 2)

Math 20 Spring 2005 Final Exam Practice Problems (Set 2) Math 2 Spring 2 Final Eam Practice Problems (Set 2) 1. Find the etreme values of f(, ) = 2 2 + 3 2 4 on the region {(, ) 2 + 2 16}. 2. Allocation of Funds: A new editor has been allotted $6, to spend on

More information

Decision tables and decision spaces

Decision tables and decision spaces Abstract Decision tables and decision spaces Z. Pawlak 1 Abstract. In this paper an Euclidean space, called a decision space is associated with ever decision table. This can be viewed as a generalization

More information

MA Game Theory 2005, LSE

MA Game Theory 2005, LSE MA. Game Theor, LSE Problem Set The problems in our third and final homework will serve two purposes. First, the will give ou practice in studing games with incomplete information. Second (unless indicated

More information

CONTINUOUS SPATIAL DATA ANALYSIS

CONTINUOUS SPATIAL DATA ANALYSIS CONTINUOUS SPATIAL DATA ANALSIS 1. Overview of Spatial Stochastic Processes The ke difference between continuous spatial data and point patterns is that there is now assumed to be a meaningful value, s

More information

( ) ( ) ( )( ) Algebra II Semester 2 Practice Exam A DRAFT. = x. Which expression is equivalent to. 1. Find the solution set of. f g x? B.

( ) ( ) ( )( ) Algebra II Semester 2 Practice Exam A DRAFT. = x. Which expression is equivalent to. 1. Find the solution set of. f g x? B. Algebra II Semester Practice Eam A. Find the solution set of 5 5 +, 9 9 +, 5 5 +, =.. Let f ( ) = and g( ) =. Which epression is equivalent to f g? ( ) ( ) + ( ) 9 9 +, 5. If h( ) = 0 and j( ) h( ) + j

More information

Chapter 9 Vocabulary Check

Chapter 9 Vocabulary Check 9 CHAPTER 9 Eponential and Logarithmic Functions Find the inverse function of each one-to-one function. See Section 9.. 67. f = + 68. f = - CONCEPT EXTENSIONS The formula = 0 e kt gives the population

More information

3.7 InveRSe FUnCTIOnS

3.7 InveRSe FUnCTIOnS CHAPTER functions learning ObjeCTIveS In this section, ou will: Verif inverse functions. Determine the domain and range of an inverse function, and restrict the domain of a function to make it one-to-one.

More information

14.1 Systems of Linear Equations in Two Variables

14.1 Systems of Linear Equations in Two Variables 86 Chapter 1 Sstems of Equations and Matrices 1.1 Sstems of Linear Equations in Two Variables Use the method of substitution to solve sstems of equations in two variables. Use the method of elimination

More information

LESSON #12 - FORMS OF A LINE COMMON CORE ALGEBRA II

LESSON #12 - FORMS OF A LINE COMMON CORE ALGEBRA II LESSON # - FORMS OF A LINE COMMON CORE ALGEBRA II Linear functions come in a variet of forms. The two shown below have been introduced in Common Core Algebra I and Common Core Geometr. TWO COMMON FORMS

More information

Computation of Csiszár s Mutual Information of Order α

Computation of Csiszár s Mutual Information of Order α Computation of Csiszár s Mutual Information of Order Damianos Karakos, Sanjeev Khudanpur and Care E. Priebe Department of Electrical and Computer Engineering and Center for Language and Speech Processing

More information

10.2 The Unit Circle: Cosine and Sine

10.2 The Unit Circle: Cosine and Sine 0. The Unit Circle: Cosine and Sine 77 0. The Unit Circle: Cosine and Sine In Section 0.., we introduced circular motion and derived a formula which describes the linear velocit of an object moving on

More information

Chapter 3. Expectations

Chapter 3. Expectations pectations - Chapter. pectations.. Introduction In this chapter we define five mathematical epectations the mean variance standard deviation covariance and correlation coefficient. We appl these general

More information

x y plane is the plane in which the stresses act, yy xy xy Figure 3.5.1: non-zero stress components acting in the x y plane

x y plane is the plane in which the stresses act, yy xy xy Figure 3.5.1: non-zero stress components acting in the x y plane 3.5 Plane Stress This section is concerned with a special two-dimensional state of stress called plane stress. It is important for two reasons: () it arises in real components (particularl in thin components

More information

Review of Elementary Probability Lecture I Hamid R. Rabiee

Review of Elementary Probability Lecture I Hamid R. Rabiee Stochastic Processes Review o Elementar Probabilit Lecture I Hamid R. Rabiee Outline Histor/Philosoph Random Variables Densit/Distribution Functions Joint/Conditional Distributions Correlation Important

More information

Functions. Introduction

Functions. Introduction Functions,00 P,000 00 0 970 97 980 98 990 99 000 00 00 Figure Standard and Poor s Inde with dividends reinvested (credit "bull": modification of work b Praitno Hadinata; credit "graph": modification of

More information

x. 4. 2x 10 4x. 10 x

x. 4. 2x 10 4x. 10 x CCGPS UNIT Semester 1 COORDINATE ALGEBRA Page 1 of Reasoning with Equations and Quantities Name: Date: Understand solving equations as a process of reasoning and eplain the reasoning MCC9-1.A.REI.1 Eplain

More information

A note on the connection between the Tsallis thermodynamics and cumulative prospect theory

A note on the connection between the Tsallis thermodynamics and cumulative prospect theory A note on the connection between the Tsallis thermodynamics and cumulative prospect theory Abstract: This note presents eplicitly a strong connection between the Tsallis thermodynamics and the so-called

More information

INF Introduction to classifiction Anne Solberg

INF Introduction to classifiction Anne Solberg INF 4300 8.09.17 Introduction to classifiction Anne Solberg anne@ifi.uio.no Introduction to classification Based on handout from Pattern Recognition b Theodoridis, available after the lecture INF 4300

More information

February 11, JEL Classification: C72, D43, D44 Keywords: Discontinuous games, Bertrand game, Toehold, First- and Second- Price Auctions

February 11, JEL Classification: C72, D43, D44 Keywords: Discontinuous games, Bertrand game, Toehold, First- and Second- Price Auctions Eistence of Mied Strateg Equilibria in a Class of Discontinuous Games with Unbounded Strateg Sets. Gian Luigi Albano and Aleander Matros Department of Economics and ELSE Universit College London Februar

More information

1.5. Analyzing Graphs of Functions. The Graph of a Function. What you should learn. Why you should learn it. 54 Chapter 1 Functions and Their Graphs

1.5. Analyzing Graphs of Functions. The Graph of a Function. What you should learn. Why you should learn it. 54 Chapter 1 Functions and Their Graphs 0_005.qd /7/05 8: AM Page 5 5 Chapter Functions and Their Graphs.5 Analzing Graphs of Functions What ou should learn Use the Vertical Line Test for functions. Find the zeros of functions. Determine intervals

More information

1.3 LIMITS AT INFINITY; END BEHAVIOR OF A FUNCTION

1.3 LIMITS AT INFINITY; END BEHAVIOR OF A FUNCTION . Limits at Infinit; End Behavior of a Function 89. LIMITS AT INFINITY; END BEHAVIOR OF A FUNCTION Up to now we have been concerned with its that describe the behavior of a function f) as approaches some

More information

Chapter 4 Analytic Trigonometry

Chapter 4 Analytic Trigonometry Analtic Trigonometr Chapter Analtic Trigonometr Inverse Trigonometric Functions The trigonometric functions act as an operator on the variable (angle, resulting in an output value Suppose this process

More information

8.1 Exponents and Roots

8.1 Exponents and Roots Section 8. Eponents and Roots 75 8. Eponents and Roots Before defining the net famil of functions, the eponential functions, we will need to discuss eponent notation in detail. As we shall see, eponents

More information

6. Vector Random Variables

6. Vector Random Variables 6. Vector Random Variables In the previous chapter we presented methods for dealing with two random variables. In this chapter we etend these methods to the case of n random variables in the following

More information

Trigonometric. equations. Topic: Periodic functions and applications. Simple trigonometric. equations. Equations using radians Further trigonometric

Trigonometric. equations. Topic: Periodic functions and applications. Simple trigonometric. equations. Equations using radians Further trigonometric Trigonometric equations 6 sllabusref eferenceence Topic: Periodic functions and applications In this cha 6A 6B 6C 6D 6E chapter Simple trigonometric equations Equations using radians Further trigonometric

More information

Definitions and Proofs

Definitions and Proofs Giving Advice vs. Making Decisions: Transparency, Information, and Delegation Online Appendix A Definitions and Proofs A. The Informational Environment The set of states of nature is denoted by = [, ],

More information

We have examined power functions like f (x) = x 2. Interchanging x

We have examined power functions like f (x) = x 2. Interchanging x CHAPTER 5 Eponential and Logarithmic Functions We have eamined power functions like f =. Interchanging and ields a different function f =. This new function is radicall different from a power function

More information

Derivatives of Multivariable Functions

Derivatives of Multivariable Functions Chapter 0 Derivatives of Multivariable Functions 0. Limits Motivating Questions In this section, we strive to understand the ideas generated b the following important questions: What do we mean b the limit

More information

Answer Explanations. The SAT Subject Tests. Mathematics Level 1 & 2 TO PRACTICE QUESTIONS FROM THE SAT SUBJECT TESTS STUDENT GUIDE

Answer Explanations. The SAT Subject Tests. Mathematics Level 1 & 2 TO PRACTICE QUESTIONS FROM THE SAT SUBJECT TESTS STUDENT GUIDE The SAT Subject Tests Answer Eplanations TO PRACTICE QUESTIONS FROM THE SAT SUBJECT TESTS STUDENT GUIDE Mathematics Level & Visit sat.org/stpractice to get more practice and stud tips for the Subject Test

More information

All parabolas through three non-collinear points

All parabolas through three non-collinear points ALL PARABOLAS THROUGH THREE NON-COLLINEAR POINTS 03 All parabolas through three non-collinear points STANLEY R. HUDDY and MICHAEL A. JONES If no two of three non-collinear points share the same -coordinate,

More information

Functions of Several Variables

Functions of Several Variables Chapter 1 Functions of Several Variables 1.1 Introduction A real valued function of n variables is a function f : R, where the domain is a subset of R n. So: for each ( 1,,..., n ) in, the value of f is

More information

Mathematical Statistics. Gregg Waterman Oregon Institute of Technology

Mathematical Statistics. Gregg Waterman Oregon Institute of Technology Mathematical Statistics Gregg Waterman Oregon Institute of Technolog c Gregg Waterman This work is licensed under the Creative Commons Attribution. International license. The essence of the license is

More information

Logarithms. Bacteria like Staph aureus are very common.

Logarithms. Bacteria like Staph aureus are very common. UNIT 10 Eponentials and Logarithms Bacteria like Staph aureus are ver common. Copright 009, K1 Inc. All rights reserved. This material ma not be reproduced in whole or in part, including illustrations,

More information

CSE 546 Midterm Exam, Fall 2014

CSE 546 Midterm Exam, Fall 2014 CSE 546 Midterm Eam, Fall 2014 1. Personal info: Name: UW NetID: Student ID: 2. There should be 14 numbered pages in this eam (including this cover sheet). 3. You can use an material ou brought: an book,

More information

Math 214 Spring problem set (a) Consider these two first order equations. (I) dy dx = x + 1 dy

Math 214 Spring problem set (a) Consider these two first order equations. (I) dy dx = x + 1 dy Math 4 Spring 08 problem set. (a) Consider these two first order equations. (I) d d = + d (II) d = Below are four direction fields. Match the differential equations above to their direction fields. Provide

More information

Exponential and Logarithmic Functions

Exponential and Logarithmic Functions Eponential and Logarithmic Functions 6 Figure Electron micrograph of E. Coli bacteria (credit: Mattosaurus, Wikimedia Commons) CHAPTER OUTLINE 6. Eponential Functions 6. Logarithmic Properties 6. Graphs

More information

1 + x 1/2. b) For what values of k is g a quasi-concave function? For what values of k is g a concave function? Explain your answers.

1 + x 1/2. b) For what values of k is g a quasi-concave function? For what values of k is g a concave function? Explain your answers. Questions and Answers from Econ 0A Final: Fall 008 I have gone to some trouble to explain the answers to all of these questions, because I think that there is much to be learned b working through them

More information

Chapter 6. Self-Adjusting Data Structures

Chapter 6. Self-Adjusting Data Structures Chapter 6 Self-Adjusting Data Structures Chapter 5 describes a data structure that is able to achieve an epected quer time that is proportional to the entrop of the quer distribution. The most interesting

More information

STUDY KNOWHOW PROGRAM STUDY AND LEARNING CENTRE. Functions & Graphs

STUDY KNOWHOW PROGRAM STUDY AND LEARNING CENTRE. Functions & Graphs STUDY KNOWHOW PROGRAM STUDY AND LEARNING CENTRE Functions & Graphs Contents Functions and Relations... 1 Interval Notation... 3 Graphs: Linear Functions... 5 Lines and Gradients... 7 Graphs: Quadratic

More information

INF Introduction to classifiction Anne Solberg Based on Chapter 2 ( ) in Duda and Hart: Pattern Classification

INF Introduction to classifiction Anne Solberg Based on Chapter 2 ( ) in Duda and Hart: Pattern Classification INF 4300 151014 Introduction to classifiction Anne Solberg anne@ifiuiono Based on Chapter 1-6 in Duda and Hart: Pattern Classification 151014 INF 4300 1 Introduction to classification One of the most challenging

More information

MATH 021 UNIT 1 HOMEWORK ASSIGNMENTS

MATH 021 UNIT 1 HOMEWORK ASSIGNMENTS MATH 01 UNIT 1 HOMEWORK ASSIGNMENTS General Instructions You will notice that most of the homework assignments for a section have more than one part. Usuall, the part (A) questions ask for eplanations,

More information

y = f(x + 4) a) Example: A repeating X by using two linear equations y = ±x. b) Example: y = f(x - 3). The translation is

y = f(x + 4) a) Example: A repeating X by using two linear equations y = ±x. b) Example: y = f(x - 3). The translation is Answers Chapter Function Transformations. Horizontal and Vertical Translations, pages to. a h, k h, k - c h -, k d h 7, k - e h -, k. a A (-,, B (-,, C (-,, D (,, E (, A (-, -, B (-,, C (,, D (, -, E (,

More information

A Bayesian Approach to Learning Causal Networks

A Bayesian Approach to Learning Causal Networks A Baesian Approach to Learning Causal Networks David Heckerman Microsoft Research, Bldg 9S/1 Redmond 98052-6399, WA heckerma@microsoft.com Abstract Whereas acausal Baesian networks represent probabilistic

More information

Section 3.1. ; X = (0, 1]. (i) f : R R R, f (x, y) = x y

Section 3.1. ; X = (0, 1]. (i) f : R R R, f (x, y) = x y Paul J. Bruillard MATH 0.970 Problem Set 6 An Introduction to Abstract Mathematics R. Bond and W. Keane Section 3.1: 3b,c,e,i, 4bd, 6, 9, 15, 16, 18c,e, 19a, 0, 1b Section 3.: 1f,i, e, 6, 1e,f,h, 13e,

More information

Functions. Introduction

Functions. Introduction Functions,00 P,000 00 0 970 97 980 98 990 99 000 00 00 Figure Standard and Poor s Inde with dividends reinvested (credit "bull": modification of work b Praitno Hadinata; credit "graph": modification of

More information

Toda s Theorem: PH P #P

Toda s Theorem: PH P #P CS254: Computational Compleit Theor Prof. Luca Trevisan Final Project Ananth Raghunathan 1 Introduction Toda s Theorem: PH P #P The class NP captures the difficult of finding certificates. However, in

More information

KEY IDEAS. Chapter 1 Function Transformations. 1.1 Horizontal and Vertical Translations Pre-Calculus 12 Student Workbook MHR 1

KEY IDEAS. Chapter 1 Function Transformations. 1.1 Horizontal and Vertical Translations Pre-Calculus 12 Student Workbook MHR 1 Chapter Function Transformations. Horizontal and Vertical Translations A translation can move the graph of a function up or down (vertical translation) and right or left (horizontal translation). A translation

More information

Computation of Total Capacity for Discrete Memoryless Multiple-Access Channels

Computation of Total Capacity for Discrete Memoryless Multiple-Access Channels IEEE TRANSACTIONS ON INFORATION THEORY, VOL. 50, NO. 11, NOVEBER 2004 2779 Computation of Total Capacit for Discrete emorless ultiple-access Channels ohammad Rezaeian, ember, IEEE, and Ale Grant, Senior

More information

SF2972 Game Theory Problem set on extensive form games

SF2972 Game Theory Problem set on extensive form games SF2972 Game Theor Problem set on etensive form games Mark Voorneveld There are five eercises, to be handed in at the final lecture (March 0). For a bonus point, please answer all questions; at least half

More information

Review of Probability

Review of Probability Review of robabilit robabilit Theor: Man techniques in speech processing require the manipulation of probabilities and statistics. The two principal application areas we will encounter are: Statistical

More information

On Information and Sufficiency

On Information and Sufficiency On Information and Sufficienc Huaiu hu SFI WORKING PAPER: 997-02-04 SFI Working Papers contain accounts of scientific work of the author(s) and do not necessaril represent the views of the Santa Fe Institute.

More information

Math Final Review. 1. Match the following functions with the given graphs without using your calculator: f3 (x) = x4 x 5.

Math Final Review. 1. Match the following functions with the given graphs without using your calculator: f3 (x) = x4 x 5. Mat 5 Final Review. Matc te following functions wit te given graps witout using our calculator: f () = /3 f4 () = f () = /3 54 5 + 5 f5 () = f3 () = 4 5 53 5 + 5 f6 () = 5 5 + 5 (Ans: A, E, D, F, B, C)

More information

The Entropy Power Inequality and Mrs. Gerber s Lemma for Groups of order 2 n

The Entropy Power Inequality and Mrs. Gerber s Lemma for Groups of order 2 n The Entrop Power Inequalit and Mrs. Gerber s Lemma for Groups of order 2 n Varun Jog EECS, UC Berkele Berkele, CA-94720 Email: varunjog@eecs.berkele.edu Venkat Anantharam EECS, UC Berkele Berkele, CA-94720

More information

Functions. Introduction CHAPTER OUTLINE

Functions. Introduction CHAPTER OUTLINE Functions,00 P,000 00 0 970 97 980 98 990 99 000 00 00 Figure Standard and Poor s Inde with dividends reinvested (credit "bull": modification of work b Praitno Hadinata; credit "graph": modification of

More information

Additional Material On Recursive Sequences

Additional Material On Recursive Sequences Penn State Altoona MATH 141 Additional Material On Recursive Sequences 1. Graphical Analsis Cobweb Diagrams Consider a generic recursive sequence { an+1 = f(a n ), n = 1,, 3,..., = Given initial value.

More information

Linear programming: Theory

Linear programming: Theory Division of the Humanities and Social Sciences Ec 181 KC Border Convex Analsis and Economic Theor Winter 2018 Topic 28: Linear programming: Theor 28.1 The saddlepoint theorem for linear programming The

More information

NEXT-GENERATION Advanced Algebra and Functions

NEXT-GENERATION Advanced Algebra and Functions NEXT-GENERATIN Advanced Algebra and Functions Sample Questions The College Board The College Board is a mission-driven not-for-profit organization that connects students to college success and opportunit.

More information

Chapter 13. Overview. The Quadratic Formula. Overview. The Quadratic Formula. The Quadratic Formula. Lewinter & Widulski 1. The Quadratic Formula

Chapter 13. Overview. The Quadratic Formula. Overview. The Quadratic Formula. The Quadratic Formula. Lewinter & Widulski 1. The Quadratic Formula Chapter 13 Overview Some More Math Before You Go The Quadratic Formula The iscriminant Multiplication of Binomials F.O.I.L. Factoring Zero factor propert Graphing Parabolas The Ais of Smmetr, Verte and

More information

Algebra 2 Unit 2 Practice

Algebra 2 Unit 2 Practice Algebra Unit Practice LESSON 7-1 1. Consider a rectangle that has a perimeter of 80 cm. a. Write a function A(l) that represents the area of the rectangle with length l.. A rectangle has a perimeter of

More information

INTRODUCTION TO DIFFERENTIAL EQUATIONS

INTRODUCTION TO DIFFERENTIAL EQUATIONS INTRODUCTION TO DIFFERENTIAL EQUATIONS. Definitions and Terminolog. Initial-Value Problems.3 Differential Equations as Mathematical Models CHAPTER IN REVIEW The words differential and equations certainl

More information

Quick Review 4.1 (For help, go to Sections 1.2, 2.1, 3.5, and 3.6.)

Quick Review 4.1 (For help, go to Sections 1.2, 2.1, 3.5, and 3.6.) Section 4. Etreme Values of Functions 93 EXPLORATION Finding Etreme Values Let f,.. Determine graphicall the etreme values of f and where the occur. Find f at these values of.. Graph f and f or NDER f,,

More information

11.4 Polar Coordinates

11.4 Polar Coordinates 11. Polar Coordinates 917 11. Polar Coordinates In Section 1.1, we introduced the Cartesian coordinates of a point in the plane as a means of assigning ordered pairs of numbers to points in the plane.

More information

4 Strain true strain engineering strain plane strain strain transformation formulae

4 Strain true strain engineering strain plane strain strain transformation formulae 4 Strain The concept of strain is introduced in this Chapter. The approimation to the true strain of the engineering strain is discussed. The practical case of two dimensional plane strain is discussed,

More information

0.24 adults 2. (c) Prove that, regardless of the possible values of and, the covariance between X and Y is equal to zero. Show all work.

0.24 adults 2. (c) Prove that, regardless of the possible values of and, the covariance between X and Y is equal to zero. Show all work. 1 A socioeconomic stud analzes two discrete random variables in a certain population of households = number of adult residents and = number of child residents It is found that their joint probabilit mass

More information

A Perspective on the Role of Engineering Decision-Based Design: Challenges and Opportunities

A Perspective on the Role of Engineering Decision-Based Design: Challenges and Opportunities A Perspective on the Role of Engineering Decision-Based Design: Challenges and Opportunities Zissimos P. Mourelatos Associate Professor Mechanical Engineering Department Oakland University Rochester, MI

More information

6.896 Topics in Algorithmic Game Theory February 8, Lecture 2

6.896 Topics in Algorithmic Game Theory February 8, Lecture 2 6.896 Topics in Algorithmic Game Theor Februar 8, 2010 Lecture 2 Lecturer: Constantinos Daskalakis Scribe: Yang Cai, Debmala Panigrahi In this lecture, we focus on two-plaer zero-sum games. Our goal is

More information

5A Exponential functions

5A Exponential functions Chapter 5 5 Eponential and logarithmic functions bjectives To graph eponential and logarithmic functions and transformations of these functions. To introduce Euler s number e. To revise the inde and logarithm

More information

11.1 Double Riemann Sums and Double Integrals over Rectangles

11.1 Double Riemann Sums and Double Integrals over Rectangles Chapter 11 Multiple Integrals 11.1 ouble Riemann Sums and ouble Integrals over Rectangles Motivating Questions In this section, we strive to understand the ideas generated b the following important questions:

More information

9-1. The Function with Equation y = ax 2. Vocabulary. Graphing y = x 2. Lesson

9-1. The Function with Equation y = ax 2. Vocabulary. Graphing y = x 2. Lesson Chapter 9 Lesson 9-1 The Function with Equation = a BIG IDEA The graph of an quadratic function with equation = a, with a 0, is a parabola with verte at the origin. Vocabular parabola refl ection-smmetric

More information

Estimation of Mixed Exponentiated Weibull Parameters in Life Testing

Estimation of Mixed Exponentiated Weibull Parameters in Life Testing International Mathematical Forum, Vol., 6, no. 6, 769-78 HIKARI Ltd, www.m-hikari.com http://d.doi.org/.988/imf.6.6445 Estimation of Mied Eponentiated Weibull Parameters in Life Testing A. M. Abd-Elrahman

More information

Conic Sections CHAPTER OUTLINE. The Circle Ellipses and Hyperbolas Second-Degree Inequalities and Nonlinear Systems FIGURE 1

Conic Sections CHAPTER OUTLINE. The Circle Ellipses and Hyperbolas Second-Degree Inequalities and Nonlinear Systems FIGURE 1 088_0_p676-7 /7/0 :5 PM Page 676 (FPG International / Telegraph Colour Librar) Conic Sections CHAPTER OUTLINE. The Circle. Ellipses and Hperbolas.3 Second-Degree Inequalities and Nonlinear Sstems O ne

More information

Strain Transformation and Rosette Gage Theory

Strain Transformation and Rosette Gage Theory Strain Transformation and Rosette Gage Theor It is often desired to measure the full state of strain on the surface of a part, that is to measure not onl the two etensional strains, and, but also the shear

More information

Sample-based Optimal Transport and Barycenter Problems

Sample-based Optimal Transport and Barycenter Problems Sample-based Optimal Transport and Barcenter Problems MAX UANG New York Universit, Courant Institute of Mathematical Sciences AND ESTEBAN G. TABA New York Universit, Courant Institute of Mathematical Sciences

More information

c) domain {x R, x 3}, range {y R}

c) domain {x R, x 3}, range {y R} Answers Chapter 1 Functions 1.1 Functions, Domain, and Range 1. a) Yes, no vertical line will pass through more than one point. b) No, an vertical line between = 6 and = 6 will pass through two points..

More information

Symmetry Arguments and the Role They Play in Using Gauss Law

Symmetry Arguments and the Role They Play in Using Gauss Law Smmetr Arguments and the Role The la in Using Gauss Law K. M. Westerberg (9/2005) Smmetr plas a ver important role in science in general, and phsics in particular. Arguments based on smmetr can often simplif

More information

Cubic and quartic functions

Cubic and quartic functions 3 Cubic and quartic functions 3A Epanding 3B Long division of polnomials 3C Polnomial values 3D The remainder and factor theorems 3E Factorising polnomials 3F Sum and difference of two cubes 3G Solving

More information

Great Expectations. Part I: On the Customizability of Generalized Expected Utility*

Great Expectations. Part I: On the Customizability of Generalized Expected Utility* Great Expectations. Part I: On the Customizability of Generalized Expected Utility* Francis C. Chu and Joseph Y. Halpern Department of Computer Science Cornell University Ithaca, NY 14853, U.S.A. Email:

More information

Correlation analysis 2: Measures of correlation

Correlation analysis 2: Measures of correlation Correlation analsis 2: Measures of correlation Ran Tibshirani Data Mining: 36-462/36-662 Februar 19 2013 1 Review: correlation Pearson s correlation is a measure of linear association In the population:

More information

y = 3 2 x 3. The slope of this line is 3 and its y-intercept is (0, 3). For every two units to the right, the line rises three units vertically.

y = 3 2 x 3. The slope of this line is 3 and its y-intercept is (0, 3). For every two units to the right, the line rises three units vertically. Mat 2 - Calculus for Management and Social Science. Understanding te basics of lines in te -plane is crucial to te stud of calculus. Notes Recall tat te and -intercepts of a line are were te line meets

More information

An outline of the Bayesian decision theory

An outline of the Bayesian decision theory An outline of the Baesian decision theor H. R. N. van Erp, R. O. Linger, and P. H. A. J. M. van Gelder Citation: AIP Conference Proceedings 1757, 050001 (016); doi: 10.1063/1.4959057 View online: https://doi.org/10.1063/1.4959057

More information

P.4 Lines in the Plane

P.4 Lines in the Plane 28 CHAPTER P Prerequisites P.4 Lines in the Plane What ou ll learn about Slope of a Line Point-Slope Form Equation of a Line Slope-Intercept Form Equation of a Line Graphing Linear Equations in Two Variables

More information

1.6 ELECTRONIC STRUCTURE OF THE HYDROGEN ATOM

1.6 ELECTRONIC STRUCTURE OF THE HYDROGEN ATOM 1.6 ELECTRONIC STRUCTURE OF THE HYDROGEN ATOM 23 How does this wave-particle dualit require us to alter our thinking about the electron? In our everda lives, we re accustomed to a deterministic world.

More information

Constructing Multiattribute Utility Functions for Decision Analysis

Constructing Multiattribute Utility Functions for Decision Analysis INFORMS 2010 c 2010 INFORMS isbn 978-0-9843378-0-4 doi 10.1287/educ.1100.0070 Constructing Multiattribute Utility Functions for Decision Analysis AliE.Abbas Department of Industrial and Enterprise Systems

More information

Infinite Limits. Let f be the function given by. f x 3 x 2.

Infinite Limits. Let f be the function given by. f x 3 x 2. 0_005.qd //0 :07 PM Page 8 SECTION.5 Infinite Limits 8, as Section.5, as + f() = f increases and decreases without bound as approaches. Figure.9 Infinite Limits Determine infinite its from the left and

More information

Noncommuting Rotation and Angular Momentum Operators

Noncommuting Rotation and Angular Momentum Operators Noncommuting Rotation and Angular Momentum Operators Originall appeared at: http://behindtheguesses.blogspot.com/2009/08/noncommuting-rotation-and-angular.html Eli Lanse elanse@gmail.com August 31, 2009

More information