Preference Handling in Relational Query Languages

1. Situation The Problem The Solution Contributions Preference Handling in Relational Query Languages Radim Nedbal Czech Technical University in Prague, Fakulty of Nuclear Sciences and Physical Engineering Prague, 7th October 2011 1,/,30

2. Situation The Problem The Solution Contributions Preference Handling in Relational Query Languages Radim Nedbal Czech Technical University in Prague, Fakulty of Nuclear Sciences and Physical Engineering Prague, 7th October 2011 1,/,30

3. Situation The Problem The Solution Contributions Contents 1 Situation Autonomous systems should make desirable choices Desirable choices can be intensionally denoted in a RQL 2 The Problem Desirable feasible choices can’t be denoted in a RQL Manual selection is opaque to the system 3 The Solution A declarative language for preferences conditional on the context represented as a relational DB instance Specifying and interpreting preferences Retrieving the most desirable choices 4 Contributions Summary and conclusions Related work 2,/,30

7. Situation The Problem The Solution Contributions Autonomous systems should make choices most desirable at the current context Complex autonomous systems (CAS) CAS Environment

8. Situation The Problem The Solution Contributions Autonomous systems should make choices most desirable at the current context Complex autonomous systems (CAS) CAS rcepts Pe Environment

9. Situation The Problem The Solution Contributions Autonomous systems should make choices most desirable at the current context Complex autonomous systems (CAS) CAS Decision m os rcepts ac t desira tions Pe bl e Environment A framework for selecting the most d e s i r a b l e f e a s i b l e c h o i c e s at run-time declarative speciﬁcation of (designer’s) d e s i r e s, amenable to customization by allowing speciﬁcation of additional d e s i r e s, by providing additional information about the c o n t e x t. 3,/,30

10. Situation The Problem The Solution Contributions Autonomous systems should make choices most desirable at the current context Complex autonomous systems (CAS) CAS Decision m os rcepts ac t desira tions Pe bl e Environment A framework for selecting the most d e s i r a b l e f e a s i b l e c h o i c e s at run-time declarative speciﬁcation of (designer’s) d e s i r e s, amenable to customization by allowing speciﬁcation of additional d e s i r e s, by providing additional information about the c o n t e x t. 3,/,30

11. Situation The Problem The Solution Contributions Autonomous systems should make choices most desirable at the current context System conﬁguration & design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA A A1 A N 0.2 0 mB B A2 A N 0.2 0 mC C A3 A N 1 1 . . . . A4 A N 1 1 . . A5 A Y 0.3 0.04 MAP mA A5 A3 A4 A1 A2 4,/,30

12. Situation The Problem The Solution Contributions Autonomous systems should make choices most desirable at the current context System conﬁguration & design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA s1 mA A A1 A N 0.2 0 mB B A2 A N 0.2 0 mC C A3 A N 1 1 . . . . A4 A N 1 1 . . A5 A Y 0.3 0.04 MAP mA A5 A3 A4 A1 A2 4,/,30

13. Situation The Problem The Solution Contributions Autonomous systems should make choices most desirable at the current context System conﬁguration & design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA s1 mA A A1 A N 0.2 0 A3 s2 mB B A2 A N 0.2 0 A4 s2 mC C A3 A N 1 1 . . . . A4 A N 1 1 . . A5 A Y 0.3 0.04 MAP mA A5 A3 A4 A1 A2 4,/,30

14. Situation The Problem The Solution Contributions Autonomous systems should make choices most desirable at the current context System conﬁguration & design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA s1 mA A A1 A N 0.2 0 A3 s2 mB B A2 A N 0.2 0 A4 s2 mC C A3 A N 1 1 A5 s2 . . . . A4 A N 1 1 . . A5 A Y 0.3 0.04 MAP mA A5 A3 A4 A1 A2 4,/,30

15. Situation The Problem The Solution Contributions Autonomous systems should make choices most desirable at the current context System conﬁguration & design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA A A1 A N 0 0 mB B A2 A N 0 0 mC C A3 A N 0 1 . . . . A4 A N 0 1 . . A5 A Y 0 0.04 MAP mA A5 A3 A4 A1 A2 4,/,30

16. Situation The Problem The Solution Contributions Autonomous systems should make choices most desirable at the current context System conﬁguration & design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE A5 s1 mA A A1 A N 0 0 A3 s2 mB B A2 A N 0 0 mC C A3 A N 0 1 . . . . A4 A N 0 1 . . A5 A Y 0 0.04 MAP mA A5 A3 A4 A1 A2 4,/,30

17. Situation The Problem The Solution Contributions Autonomous systems should make choices most desirable at the current context System conﬁguration & design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA A A1 A N 1 0 mB B A2 A N 1 0 mC C A3 A N 1 1 . . . . A4 A N 1 1 . . A5 A Y 1 0.04 MAP mA A5 A3 A4 A1 A2 4,/,30

18. Situation The Problem The Solution Contributions Autonomous systems should make choices most desirable at the current context System conﬁguration & design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA s1 mA A A1 A N 1 0 mB B A2 A N 1 0 mC C A3 A N 1 1 . . . . A4 A N 1 1 . . A5 A Y 1 0.04 MAP mA A5 A3 A4 A1 A2 4,/,30

19. Situation The Problem The Solution Contributions Autonomous systems should make choices most desirable at the current context System conﬁguration & design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA s1 mA A A1 A N 1 0 A3 s2 mB B A2 A N 1 0 A4 s2 mC C A3 A N 1 1 . . . . A4 A N 1 1 . . A5 A Y 1 0.04 MAP mA A5 A3 A4 A1 A2 4,/,30

20. Situation The Problem The Solution Contributions Autonomous systems should make choices most desirable at the current context System conﬁguration & design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA s1 mA A A1 A N 1 0 A3 s2 mB B A2 A N 1 0 A4 s2 mC C A3 A N 1 1 A1 s2 . . . . A4 A N 1 1 . . A2 s2 A5 A Y 1 0.04 MAP mA A5 A3 A4 A1 A2 4,/,30

21. Situation The Problem The Solution Contributions Desirable choices can be intensionally denoted by their properties in a RQL DB of feasible choices INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA s1 mA A . . . . . . . . . . . . . . . . . . . mB B . . A4 A N 1 1 mC C A5 A Y 0.3 0.04 . . . . . . B1 B N 0 1 . . . . . . . . . . . . . . . Maps of rooms where some non-IR cameras shoot a lit area R(xmap , s1) ←− S(xmap , xroom ) ∧ T (xcamera , xroom , “N”, 1, xgate ) R( mA, s1) ←− S( mA , A ) ∧ T ( A4 , A , “N”, 1, 1 ) A DB query 1 is system interpretable speciﬁcation of desirable choices, 2 can be re-evaluated when DB changes. 5,/,30

22. Situation The Problem The Solution Contributions Desirable choices can be intensionally denoted by their properties in a RQL DB of feasible choices INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA s1 mA A . . . . . . . . . . . . . . . . . . . mB B . . A4 A N 1 1 mC C A5 A Y 0.3 0.04 . . . . . . B1 B N 0 1 . . . . . . . . . . . . . . . Maps of rooms where some non-IR cameras shoot a lit area R(xmap , s1) ←− S(xmap , xroom ) ∧ T (xcamera , xroom , “N”, 1, xgate ) R( mA, s1) ←− S( mA , A ) ∧ T ( A4 , A , “N”, 1, 1 ) A DB query 1 is system interpretable speciﬁcation of desirable choices, 2 can be re-evaluated when DB changes. 5,/,30

23. Situation The Problem The Solution Contributions Desirable choices can be intensionally denoted by their properties in a RQL A DB query speciﬁes desirable characteristics Non-IR cameras shooting a lit gate area. ans(xcamera ) ←− T (xcamera , xroom , xIR , xlit , xgate ) ∧ xIR = “N” ∧ xlit = 1 ∧ xgate = 1 . T : relation of installed cameras, x IR = “N” : non-IR cameras, x lit = 1 : cameras shooting a lit area, x gate = 1 : cameras shooting a gate area. (Most) desirable feasible choices are what matters (most)! 6,/,30

28. Situation The Problem The Solution Contributions (Most) desirable feasible choices can’t be intensionally denoted by their properties in a RQL Little knowledge to specify characteristics of feasible choices Asking too speciﬁcally empty result effect. (Satisﬁability of DB queries is undecidable)

29. Adjust characteristics or give up! Asking for too little ﬂooding effect.

30. Manual selection! Gradual adjusting original characteristics

31. Add or remove characteristics!

32. Relax or tighten up characteristics! b expensive as space of characteristics is combinatorially huge! b infeasible in the case of automated decision making (autonomous agents)!! 7,/,30

68. Situation The Problem The Solution Contributions (Most) desirable feasible choices can’t be intensionally denoted by their properties in a RQL Gradual adjusting characteristics Non-IR cameras shooting a lit gate area. x IR = “N” : non-IR cameras, x lit = 1 : cameras shooting a lit area, x gate = 1 : cameras shooting a gate area. x IR = “N” ∧ x lit = 1 ∧ x gate = 1 x IR = “N” ∧ x lit = 1 x lit = 1 ∧ x gate = 1 x IR = “N” ∧ x gate = 1 x IR = “N” x lit = 1 x gate = 1 8,/,30

76. Situation The Problem The Solution Contributions Manual selection or adjusting characteristics of desired choices is opaque to the system Reasons behind manual selection of adjusting are opaque to the system, are someone’s “liking of one thing more than another,” i.e., various desirability of respective answers, are what we term preferences. Preferences are wishes! No perfect match?? worse alternatives. A paradigm shift from exact matches towards a best possible match-making, from h a r d c o n s t r a i n t s to s o f t c o n s t r a i n t s. The main goal of the thesis a general framework for incorporating preferences in RQL to support the user-friendly design of autonomous systems that can act in dynamic environment. 9,/,30

79. Situation The Problem The Solution Contributions Manual selection or adjusting characteristics of desired choices is opaque to the system Requirements þ J 10,/,30

80. Situation The Problem The Solution Contributions Manual selection or adjusting characteristics of desired choices is opaque to the system Requirements RQL q þ J 10,/,30

81. Situation The Problem The Solution Contributions Manual selection or adjusting characteristics of desired choices is opaque to the system q(J) Requirements RQL q þ J 10,/,30

82. Situation The Problem The Solution Contributions Manual selection or adjusting characteristics of desired choices is opaque to the system nual designati Ma on Preferences q(J) Requirements RQL q þ J 10,/,30

83. Situation The Problem The Solution Contributions Manual selection or adjusting characteristics of desired choices is opaque to the system Requirements, preferences þ J 10,/,30

84. Situation The Problem The Solution Contributions Manual selection or adjusting characteristics of desired choices is opaque to the system Requirements, preferences RQL + q∗ þ J 10,/,30

85. Situation The Problem The Solution Contributions Manual selection or adjusting characteristics of desired choices is opaque to the system q ∗ (J) Requirements, preferences RQL + q∗ þ J 10,/,30

86. Situation The Problem The Solution Contributions Manual selection or adjusting characteristics of desired choices is opaque to the system nual designati Ma on Preferences q(J) P Requirements RQL q þ Back to MM Back to Representation J 10,/,30

87. Situation The Problem The Solution Contributions A declarative language for preferences conditional on the current state of the world represented as a relational DB instance Concretization of the basic concepts To J, q, P Submodels of distinguished Non-monotonic reasoning preference models Interpretation Representation DDP and DBS Models Language Algorithms Partial pre-orders Heterogenous and possibly conﬂicting preference formulae of LP Data model Query RDM The most desirable choices 11,/,30

94. Situation The Problem The Solution Contributions A declarative language for preferences conditional on the current state of the world represented as a relational DB instance Concretization of the basic concepts To J, q, P Submodels of distinguished Non-monotonic reasoning preference models Interpretation Representation DDP and DBS Models Language Algorithms A nonempty set Heterogenous and of distinguished possibly conﬂicting preference models preference formulae of LP Data model Query RDM The most desirable choices 11,/,30

99. Situation The Problem The Solution Contributions Specifying and interpreting preferences Models Back to the meta-model are structures that capture properties of speciﬁed preferences Preference model Ω, is a partial pre-order over a set Ω of a c c e p t a b l e f e a s i b l e choice. reﬂexive, transitive, partial. TAXI TAXI WALKING SUBWAY ? ? SUBWAY SUBWAY TAXI ? WALKING WALKING TAXI WALKING b Ω is abstracted as q(J); b w w (w w ) reads: “w is (strictly) preferred to w .” 12,/,30

105. Situation The Problem The Solution Contributions Specifying and interpreting preferences Language Back to the meta-model encodes preferences by specifying models Language of preference formulae LP ϕ ψ is a preference formula (of LP ) iff ϕ, ψ are DB queries “of the same type,” is represents a recognized kind of a preference. 13,/,30

106. Situation The Problem The Solution Contributions Specifying and interpreting preferences Language Back to the meta-model encodes preferences by specifying models Language of preference formulae LP ϕ ψ is a preference formula (of LP ) iff ϕ, ψ are DB queries “of the same type,” is represents a recognized kind of a preference. ϕ1 mM ψ , ϕ2 M M ψ , ϕ3 mm ψ , q(J) ϕ4 M m ψ , P .

107. Situation The Problem The Solution Contributions Specifying and interpreting preferences Language Back to the meta-model encodes preferences by specifying models Language of preference formulae LP ϕ ψ is a preference formula (of LP ) iff ϕ, ψ are DB queries “of the same type,” is represents a recognized kind of a preference. ϕ1 (J) ϕ1 mM ψ , ϕ2 M M ψ , mm ψ(J) q(J) ϕ3 ψ , ϕ4 M m ψ , P .

108. Situation The Problem The Solution Contributions Specifying and interpreting preferences Language Back to the meta-model encodes preferences by specifying models Language of preference formulae LP ϕ ψ is a preference formula (of LP ) iff ϕ, ψ are DB queries “of the same type,” is represents a recognized kind of a preference. ϕ1 (J) ϕ1 mM ψ , M M ϕ2 (J) ϕ2 ψ , ψ(J) q(J) ϕ3 mm ψ , ϕ4 M m ψ , P .

109. Situation The Problem The Solution Contributions Specifying and interpreting preferences Language Back to the meta-model encodes preferences by specifying models Language of preference formulae LP ϕ ψ is a preference formula (of LP ) iff ϕ, ψ are DB queries “of the same type,” is represents a recognized kind of a preference. ϕ1 (J) ϕ1 mM ψ , M M ϕ2 (J) ϕ2 ψ , ψ(J) q(J) ϕ3 mm ψ , ϕ4 M m ψ , ϕ3 (J) P .

110. Situation The Problem The Solution Contributions Specifying and interpreting preferences Language Back to the meta-model encodes preferences by specifying models Language of preference formulae LP ϕ ψ is a preference formula (of LP ) iff ϕ, ψ are DB queries “of the same type,” is represents a recognized kind of a preference. ϕ1 (J) ϕ1 mM ψ , M M ϕ2 (J) ϕ2 ψ , ψ(J) q(J) ϕ3 mm ψ , ϕ4 (J) ϕ4 M m ψ , ϕ3 (J) P . 13,/,30

111. Situation The Problem The Solution Contributions Specifying and interpreting preferences Language Back to the meta-model encodes preferences by specifying models Language of preference formulae LP ϕ ψ is a preference formula (of LP ) iff ϕ, ψ are DB queries “of the same type,” is represents a recognized kind of a preference. ϕ1 (J) ϕ1 mM ψ , M M ϕ2 (J) ϕ2 ψ , ψ(J) q(J) ϕ3 mm ψ , ϕ4 (J) ϕ4 M m ψ , ϕ3 (J) P . 13,/,30

112. Situation The Problem The Solution Contributions Specifying and interpreting preferences Interpretation Back to the meta-model gives exact meaning to preference formulae P = {ϕ mM ψ , ψ mM ω} ϕ ∧ ψ ∧ ¬ω ? ϕ ∧ ¬ψ q(J) q(J) 1 Minimal logic of preference:

113. w is as good as w iff allowed by P

114. each P is satisﬁed by one or more models! 2 Non-monotonic reasoning mechanism: yields DPMs.

115. Situation The Problem The Solution Contributions Specifying and interpreting preferences Interpretation Back to the meta-model gives exact meaning to preference formulae P = {ϕ mM ψ , ψ mM ω} ϕ ∧ ψ ∧ ¬ω ? ϕ ∧ ¬ψ q(J) q(J) ϕ(J) ϕ(J) ψ(J) ψ(J) 1 Minimal logic of preference:

118. Situation The Problem The Solution Contributions Specifying and interpreting preferences Interpretation Back to the meta-model gives exact meaning to preference formulae P = {ϕ mM ψ , ψ mM ω} ϕ ∧ ψ ∧ ¬ω ? ϕ ∧ ¬ψ q(J) q(J) ϕ(J) ϕ(J) ψ(J) ψ(J) ω(J) ω(J) 1 Minimal logic of preference:

123. each P is satisﬁed by one or more models! 2 Non-monotonic reasoning mechanism: yields DPMs. 14,/,30

124. Situation The Problem The Solution Contributions Specifying and interpreting preferences Interpretation Back to the meta-model gives exact meaning to preference formulae P = {ϕ mM ψ , ψ mM ω} ϕ ∧ ψ ∧ ¬ω ? ¬ϕ ∧ ψ ∧ ω q(J) q(J) ϕ(J) ϕ(J) ψ(J) ψ(J) ω(J) ω(J) 1 Minimal logic of preference:

127. Situation The Problem The Solution Contributions Specifying and interpreting preferences Interpretation Back to the meta-model gives exact meaning to preference formulae P = {ϕ mM ψ , ψ mM ω} ϕ ∧ ψ ∧ ¬ω ? ¬ϕ ∧ ψ ∧ ω q(J) q(J) ϕ(J) ϕ(J) ψ(J) ψ(J) ω(J) ω(J) 1 Minimal logic of preference:

139. Situation The Problem The Solution Contributions Retrieving the most desirable choices Representation Back to the meta-model captures preference formulae in a framework suitable for algorithms q(J) Due to Theorem 3, we can ﬁnd q (J), ϕ(J) q (J) ⊆ q(J) , ψ(J) so that the set of DPMs with underlying set q (J) determines ω(J) the set of DPMs with underlying set q(J) Any P can be represented compactly: To J, q, P the set of d i s t i n g u i s h e d p r e f e r e n c e m o d e l s,

140. deﬁning the meaning of P can be represented as the set of t h e i r s u b m o d e l s. 15,/,30

141. Situation The Problem The Solution Contributions Retrieving the most desirable choices Representation Back to the meta-model captures preference formulae in a framework suitable for algorithms q(J) Due to Theorem 3, we can ﬁnd q (J), ϕ(J) q (J) ⊆ q(J) , q (J) ψ(J) so that the set of DPMs with underlying set q (J) determines ω(J) the set of DPMs with underlying set q(J) Any P can be represented compactly: To J, q, P the set of d i s t i n g u i s h e d p r e f e r e n c e m o d e l s,

143. Situation The Problem The Solution Contributions Retrieving the most desirable choices Representation Back to the meta-model captures preference formulae in a framework suitable for algorithms q(J) Due to Theorem 3, we can ﬁnd q (J), q (J) ⊆ q(J) , q (J) so that the set of DPMs with underlying set q (J) determines the set of DPMs with underlying set q(J) Any P can be represented compactly: To J, q, P the set of d i s t i n g u i s h e d p r e f e r e n c e m o d e l s,

153. Situation The Problem The Solution Contributions Retrieving the most desirable choices Algorithms To Algorithm 2 Back to the meta-model MDC w.r.t. J, q, P = {ϕ mM ψ , ψ mM ω} ϕ(J) ψ(J) ω(J) J, q, P Decl. sem. G DPMs G MDC Theorems 1,2 s y ˆ Ùs v y Theorem 3 Theorem 5 ‚ Constr. sem. !… RQL $ DDP G Repres. G R-MDC Alg.1, Theo.4

154. Situation The Problem The Solution Contributions Retrieving the most desirable choices Algorithms To Algorithm 2 Back to the meta-model MDC w.r.t. J, q, P = {ϕ mM ψ , ψ mM ω} q (J) J, q, P Decl. sem. G DPMs G MDC Theorems 1,2 s y ˆ Ùs v y Theorem 3 Theorem 5 ‚ Constr. sem. !… RQL $ DDP G Repres. G R-MDC Alg.1, Theo.4

155. Situation The Problem The Solution Contributions Retrieving the most desirable choices Algorithms To Algorithm 2 Back to the meta-model MDC w.r.t. J, q, P = {ϕ mM ψ , ψ mM ω} b a c g f d e J, q, P Decl. sem. G DPMs G MDC Theorems 1,2 s y ˆ Ùs v y Theorem 3 Theorem 5 ‚ Constr. sem. !… RQL $ DDP G Repres. G R-MDC Alg.1, Theo.4

156. Situation The Problem The Solution Contributions Retrieving the most desirable choices Algorithms To Algorithm 2 Back to the meta-model MDC w.r.t. J, q, P = {ϕ mM ψ , ψ mM ω} ϕ ϕ mM ψ : g b∧e b b a m M ψ ω: d a∧d g ψ c g f transitivity: d e x z ∧ y ∈ {a, . . . , g} → x y ∨y z J, q, P Decl. sem. G DPMs G MDC Theorems 1,2 s y ˆ Ùs v y Theorem 3 Theorem 5 ‚ Constr. sem. !… RQL $ DDP G Repres. G R-MDC Alg.1, Theo.4

157. Situation The Problem The Solution Contributions Retrieving the most desirable choices Algorithms To Algorithm 2 Back to the meta-model MDC w.r.t. J, q, P = {ϕ mM ψ , ψ mM ω} ϕ mM ψ : g b∧e b b a m M ψ ω: d a∧d g ψ c g f transitivity: ω d e x z ∧ y ∈ {a, . . . , g} → x y ∨y z J, q, P Decl. sem. G DPMs G MDC Theorems 1,2 s y ˆ Ùs v y Theorem 3 Theorem 5 ‚ Constr. sem. !… RQL $ DDP G Repres. G R-MDC Alg.1, Theo.4

158. Situation The Problem The Solution Contributions Retrieving the most desirable choices Algorithms To Algorithm 2 Back to the meta-model MDC w.r.t. J, q, P = {ϕ mM ψ , ψ mM ω} ϕ ϕ mM ψ : g b∧e b b a m M ψ ω: d a∧d g ψ c g f transitivity: ω d e x z ∧ y ∈ {a, . . . , g} → x y ∨y z J, q, P Decl. sem. G DPMs G MDC Theorems 1,2 s y ˆ Ùs v y Theorem 3 Theorem 5 ‚ Constr. sem. !… RQL $ DDP G Repres. G R-MDC Alg.1, Theo.4

159. Situation The Problem The Solution Contributions Retrieving the most desirable choices Algorithms To Algorithm 2 Back to the meta-model MDC w.r.t. J, q, P = {ϕ mM ψ , ψ mM ω} q(J) ϕ ϕ mM ψ : g b∧e b b a m M ψ ω: d a∧d g ψ c g f transitivity: ω d e x z ∧ y ∈ {a, . . . , g} → x y ∨y z J, q, P Decl. sem. G DPMs G MDC Theorems 1,2 s y ˆ Ùs v y Theorem 3 Theorem 5 ‚ Constr. sem. !… RQL $ DDP G Repres. G R-MDC Alg.1, Theo.4

160. Situation The Problem The Solution Contributions Retrieving the most desirable choices Algorithms To Algorithm 2 Back to the meta-model MDC w.r.t. J, q, P = {ϕ mM ψ , ψ mM ω} ϕ ϕ mM ψ : g b∧e b b a m M ψ ω: d a∧d g ψ c g f transitivity: ω d e x z ∧ y ∈ {a, . . . , g} → x y ∨y z J, q, P Decl. sem. G DPMs G MDC a Theorems 1,2 s y ˆ Ùs [q∧ϕ∧ψ∧¬ω](J) v y Theorem 3 Theorem 5 ‚ Constr. sem. b !… RQL $ DDP G Repres. G R-MDC [q∧ϕ∧¬ψ∧¬ω](J) Alg.1, Theo.4

161. Situation The Problem The Solution Contributions Summary and conclusions The proposed framework is general enough to have wide applicability A novel, ﬂexible approach based on preference statements of encoding qualitative comparative the language capable 1 that may be of various kinds 2 that may be nondeterministic; 3 that may be context sensitive is suitable for control of dynamic systems, where both the 4 that may be augmented by mandatory requirements. state and number of objects changes. A camera stream of a gate area is always desirable.

162. .. an arbitrary such a camera. Streams from non-IR cameras shooting a lit area are more desirable than streams from IR cameras.

163. .. currently lit areas wrt. the updated DB. 17,/,30

176. Situation The Problem The Solution Contributions Summary and conclusions The proposed framework is formal enough to support automated decision making 1 Preferences are embedded in RQLs. 2 The empty result effect is eliminated:

177. any preference speciﬁcation has a DPM Theorem 1(totality of interpretation). 3 Constructive semantics is based on a compact representation (Theorem 3). from which DPMs can be inferred; which can be encoded as a DDP.

178. We exploit DDP machinery (Algorithm 1, Theorem 4) to compute DPMs. 4 MDC are denoted as a DB query (Theorem 5) and retrieved from the DB, exploiting standard DB optimization strategies. 18,/,30

188. Situation The Problem The Solution Contributions Related work Inﬂuential paper, projects, and ﬁgures M. Lacroix and Pierre Lavency. Preferences: Putting More Knowledge into Queries. VLDB, 1987. It’s a Preference World 1999 – Werner University of Augsburg (8 projects) Kießling Germany Preference Queries Jan 2003 – University at Buffalo Chomicki USA Command Control Ronen I. ?– Ben-Gurion University Brafman Beer-Sheva, Israel 19,/,30

189. Situation The Problem The Solution Contributions Related work Comparison to related work Preference model Language Interpretation Granularity Context Nondeterminist. Ceteris paribus Total pre-order Set of tuples Context free Strict order Totalitarian Extrinsicity Total order Pre-order Attribute External Internal Tuple Lacroix and Lavency 1987 Kießling 2002 Chomicki 2002; 2003 ( ) Holland and Kießling 2004 Brafman and Domshlak 2004 Agrawal et al. 2006 Endres and Kießling 2006 Ciaccia 2007 Mindolin and Chomicki 2007 Georgiadis et al. 2008 Kaci and Neves 2010 Zhang and Chomicki 2011 The presented approach 20,/,30

190. Appendix System conﬁguration design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA s1 mA A A1 A N 0.2 0 A3 s2 mB B A2 A N 0.2 0 mC C A3 A N 1 1 . . . . A4 A N 1 1 . . A5 A Y 0.3 0.04 MAP mA A5 A3 A4 A1 A2 23,/,30

191. Appendix System conﬁguration design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA s1 mA A A1 A N 1 0 A3 s2 mB B A2 A N 1 0 A4 s2 mC C A3 A N 1 1 A1 s2 . . . . A4 A N 1 1 . . A5 A Y 1 0.04 MAP mA A5 A3 A4 A1 A2 23,/,30

192. Appendix System conﬁguration design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA s1 mA A A1 A N 1 0 A3 s2 mB B A2 A N 1 0 A4 s2 mC C A3 A N 1 1 A2 s2 . . . . A4 A N 1 1 . . A5 A Y 1 0.04 MAP mA A5 A3 A4 A1 A2 23,/,30

193. Appendix System conﬁguration design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA s1 mA A A1 A N 0.2 0 A4 s2 mB B A2 A N 0.2 0 mC C A3 A N 1 1 . . . . A4 A N 1 1 . . A5 A Y 0.3 0.04 MAP mA A5 A3 A4 A1 A2 23,/,30

194. Appendix System conﬁguration design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA s1 mA A A1 A N 1 0 A3 s2 mB B A2 A N 1 0 mC C A3 A N 1 1 . . . . A4 A N 1 1 . . A5 A Y 1 0.04 MAP mA A5 A3 A4 A1 A2 23,/,30

195. Appendix System conﬁguration design example INPUT SCR . MAP ROOM CAMERA ROOM IR LIT GATE mA s1 mA A A1 A N 1 0 A4 s2 mB B A2 A N 1 0 mC C A3 A N 1 1 . . . . A4 A N 1 1 . . A5 A Y 1 0.04 MAP mA A5 A3 A4 A1 A2 23,/,30

196. Appendix Language Back to the meta-model encodes preferences by specifying models Language of preference formulae LP ϕ ψ is a preference formula (of LP ) iff ϕ, ψ are DB queries “of the same type,” is represents a recognized kind of a preference. ϕ1 (J) ϕ1 mM ψ , M M ϕ2 (J) ϕ2 ψ , ψ(J) q(J) ϕ3 mm ψ , ϕ4 (J) ϕ4 M m ψ , ϕ3 (J) P . 24,/,30

197. s “of the same type,” Appendix ecognized kind of a preference. Language Back to the meta-model encodes preferences by specifying models ϕ1 (J) ψ(J) q 24,/,30

198. s “of the same type,” Appendix ecognized kind of a preference. Language Back to the meta-model encodes preferences by specifying models ϕ2 (J) ψ(J) q 24,/,30

199. s “of the same type,” Appendix ecognized kind of a preference. Language Back to the meta-model encodes preferences by specifying models ψ(J) q ϕ3 (J) 24,/,30

200. s “of the same type,” Appendix ecognized kind of a preference. Language Back to the meta-model encodes preferences by specifying models ψ(J) q ϕ4 (J) 24,/,30

201. s “of the same type,” Appendix ecognized kind of a preference. Language Back to the meta-model encodes preferences by specifying models ϕ1 (J) ϕ2 (J) ψ(J) q ϕ4 (J) ϕ3 (J) 24,/,30

202. Appendix Interpretation Back to the meta-model = {ϕ mM ψ ,ψ mM gives exact meaning to preference formulae ω} ϕ∧ψ∧¬ q(J) q ϕ(J) ψ(J) 25,/,30

203. Appendix Interpretation Back to the meta-model = {ϕ mM ψ ,ψ mM gives exact meaning to preference formulae ω} ϕ∧ψ∧¬ q(J) q ψ(J) ω(J) 25,/,30

204. Appendix Interpretation Back to the meta-model ω} ϕ ∧ ψ ∧ ¬ω ? gives exact meaning to preference formulae ϕ ∧ ¬ψ q(J) ϕ(J) ψ(J) 25,/,30

205. Appendix Interpretation Back to the meta-model ω} ϕ ∧ ψ ∧ ¬ω ? gives exact meaning to preference formulae ϕ ∧ ¬ψ q(J) ψ(J) ω(J) 25,/,30

206. Appendix Interpretation Back to the meta-model = {ϕ mM ψ ,ψ mM gives exact meaning to preference formulae ω} ϕ∧ψ∧¬ q(J) q ϕ(J) ψ(J) ω(J) 25,/,30

207. Appendix Interpretation Back to the meta-model = {ϕ mM ψ ,ψ mM gives exact meaning to preference formulae ω} ϕ∧ψ∧¬ q(J) q ϕ(J) ψ(J) ω(J) 25,/,30

208. Appendix Interpretation Back to the meta-model ω} ϕ ∧ ψ ∧ ¬ω ? gives exact meaning to preference formulae ϕ ∧ ¬ψ q(J) ϕ(J) ψ(J) ω(J) 25,/,30

209. Appendix Interpretation Back to the meta-model ω} ϕ ∧ ψ ∧ ¬ω ? gives exact meaning to preference formulae ϕ ∧ ¬ψ q(J) ϕ(J) ψ(J) ω(J) 25,/,30

210. Appendix q(J) Representation Back to the meta-model captures preference formulae in a framework suitable for algorithms 26,/,30

211. Appendix Algorithms To Algorithm 2 Back to the meta-model b a c g f tr d e 27,/,30

212. Appendix Algorithms To Algorithm 2 Back to the meta-model ϕ b a ψ c g f tr d e 27,/,30

213. Appendix Algorithms To Algorithm 2 Back to the meta-model b a ψ c g f tr ω d e 27,/,30

214. Appendix Algorithms To Algorithm 2 Back to the meta-model ϕ b a ψ c g f tr ω d e 27,/,30

215. Appendix Algorithm 2 for computation of most preferred matches To algorithms Require: P, q, J. Ensure: The most preferred tuples wrt. P that fulﬁll q. 1: Construct UP . Step I. (see page 68) 2: Construct rules of P. Step II. (see page 72) 3: Add rules ensuring transitivity. Step III. (see page 74) 4: Compute O. Algorithm 1 on page 77 5: Determine O . 6: Compute SMJ O . P 7: Compute MX SMJ O P . 8: Translate q together with elements from MX SMJ O P into a RQL formula q . 9: Evaluate q (J) – the most preferred matches. DBMS 28,/,30

216. Appendix P, Ω, J _ _ _ _ _ _ _G IP (Ω, J) = UEΩ declarative P αΩ αΩ IP Ω, J ˙ semantics y “ y y y y MX (IP (Ω, J)) = Dom(wk ) y y y wk ∈MX IP (UP ,J) IP Ω, J ˙ y y y x y 8,9 1−3 y MX (IP (UP , J)) y ‰ f y y 7 y y Ð y SMJ EMJ MODP (UP , J) IP (UP , J) y P P y y y 1 c y 6 4,5 P GO EMJ MODP (UP , J) constructive P semantics 29,/,30

217. Appendix Inﬂuential paper, projects, and ﬁgures C B 30,/,30

218. Appendix Inﬂuential paper, projects, and ﬁgures Werner Kießling 30,/,30

219. Appendix Inﬂuential paper, projects, and ﬁgures Jan Chomicki 30,/,30

220. Appendix Inﬂuential paper, projects, and ﬁgures Ronen I. Brafman 30,/,30

Preference Handling in Relational Query Languages

Recommended

More Related Content

Recently uploaded (20)

Featured (20)

Preference Handling in Relational Query Languages