A Thorough Formalization of Conceptual Spaces Lucas Bechberger and - - PowerPoint PPT Presentation
A Thorough Formalization of Conceptual Spaces Lucas Bechberger and - - PowerPoint PPT Presentation
A Thorough Formalization of Conceptual Spaces Lucas Bechberger and Kai-Uwe Khnberger The Different Layers of Representation A Thorough Formalization of Conceptual Spaces / Bechberger and Khnberger 2 The Different Layers of Representation
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 2
The Different Layers of Representation
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 3
The Different Layers of Representation
Symbolic Layer ∀ x:apple (x ) ⇒red(x ) Formal Logics
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 4
The Different Layers of Representation
Symbolic Layer Subsymbolic Layer [0.42; -1.337, ...] ∀ x:apple (x ) ⇒red(x ) Formal Logics Sensor / Feature Values
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 5
The Different Layers of Representation
Symbolic Layer Subsymbolic Layer [0.42; -1.337, ...] ∀ x:apple (x ) ⇒red(x ) Formal Logics Sensor / Feature Values
?
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 6
The Different Layers of Representation
Symbolic Layer Subsymbolic Layer [0.42; -1.337, ...] ∀ x:apple (x ) ⇒red(x ) Formal Logics Sensor / Feature Values Conceptual Layer Geometric Representation
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 7
The Different Layers of Representation
Symbolic Layer Subsymbolic Layer [0.42; -1.337, ...] ∀ x:apple (x ) ⇒red(x ) Formal Logics Sensor / Feature Values Conceptual Layer Geometric Representation
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 8
The Different Layers of Representation
Symbolic Layer Subsymbolic Layer [0.42; -1.337, ...] ∀ x:apple (x ) ⇒red(x ) Formal Logics Sensor / Feature Values Conceptual Layer Geometric Representation
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 9
Conceptual Spaces [Gärdenfors2000]
[Gärdenfors2000] Gärdenfors, P. Conceptual Spaces: The Geometry of Thought. MIT press, 2000
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 10
Conceptual Spaces [Gärdenfors2000]
- Quality dimensions
- Interpretable ways of judging the similarity of two instances
- E.g., temperature, weight, brightness, pitch
[Gärdenfors2000] Gärdenfors, P. Conceptual Spaces: The Geometry of Thought. MIT press, 2000
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 11
Conceptual Spaces [Gärdenfors2000]
- Quality dimensions
- Interpretable ways of judging the similarity of two instances
- E.g., temperature, weight, brightness, pitch
- Domain
- Set of dimensions that inherently belong together
- Color: hue, saturation, and brightness
[Gärdenfors2000] Gärdenfors, P. Conceptual Spaces: The Geometry of Thought. MIT press, 2000
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 12
Conceptual Spaces [Gärdenfors2000]
- Quality dimensions
- Interpretable ways of judging the similarity of two instances
- E.g., temperature, weight, brightness, pitch
- Domain
- Set of dimensions that inherently belong together
- Color: hue, saturation, and brightness
- Distance in this space is inversely related to similarity
- Within a domain: Euclidean distance
- Between domains: Manhattan distance
[Gärdenfors2000] Gärdenfors, P. Conceptual Spaces: The Geometry of Thought. MIT press, 2000
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 13
Conceptual Spaces [Gärdenfors2000]
- Quality dimensions
- Interpretable ways of judging the similarity of two instances
- E.g., temperature, weight, brightness, pitch
- Domain
- Set of dimensions that inherently belong together
- Color: hue, saturation, and brightness
- Distance in this space is inversely related to similarity
- Within a domain: Euclidean distance
- Between domains: Manhattan distance
- Concepts
- Region + correlation information + salience weights
[Gärdenfors2000] Gärdenfors, P. Conceptual Spaces: The Geometry of Thought. MIT press, 2000
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 14
Betweenness
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 15
Betweenness
- B(x,y,z) :↔ d(x,y) + d(y,z) = d(x,z)
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 16
Betweenness
- B(x,y,z) :↔ d(x,y) + d(y,z) = d(x,z)
https://en.wikipedia.org/wiki/Taxicab_geometry#/ media/File:Manhattan_distance.svg
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 17
Betweenness
- B(x,y,z) :↔ d(x,y) + d(y,z) = d(x,z)
https://en.wikipedia.org/wiki/Taxicab_geometry#/ media/File:Manhattan_distance.svg
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 18
Betweenness
- B(x,y,z) :↔ d(x,y) + d(y,z) = d(x,z)
https://en.wikipedia.org/wiki/Taxicab_geometry#/ media/File:Manhattan_distance.svg
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 19
Betweenness
- B(x,y,z) :↔ d(x,y) + d(y,z) = d(x,z)
https://en.wikipedia.org/wiki/Taxicab_geometry#/ media/File:Manhattan_distance.svg
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 20
Betweenness
- B(x,y,z) :↔ d(x,y) + d(y,z) = d(x,z)
- Convex region C:
https://en.wikipedia.org/wiki/Taxicab_geometry#/ media/File:Manhattan_distance.svg
∀ x,z∈C :∀ y: B ( x,y,z)⇒ y∈C
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 21
Betweenness
- B(x,y,z) :↔ d(x,y) + d(y,z) = d(x,z)
- Convex region C:
- Star-shaped region S w.r.t. p:
https://en.wikipedia.org/wiki/Taxicab_geometry#/ media/File:Manhattan_distance.svg
∀ x,z∈C :∀ y: B ( x,y,z)⇒ y∈C ∀ z∈S: ∀ y : B (p,y,z)⇒ y∈S
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 22
Convexity and Manhattan distance
height age
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 23
Convexity and Manhattan distance
height age adult child
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 24
Convexity and Manhattan distance
height age adult child
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 25
Convexity and Manhattan distance
height age adult child
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 26
Convexity and Manhattan distance
height age adult child
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 27
Convexity and Manhattan distance
height age adult child
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 28
Convexity and Manhattan distance
height age adult child
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 29
Formalizing Star-Shaped Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 30
Formalizing Star-Shaped Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 31
Formalizing Star-Shaped Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 32
Formalizing Star-Shaped Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 33
Formalizing Star-Shaped Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 34
Formalizing Star-Shaped Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 35
Formalizing Star-Shaped Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 36
Formalizing Star-Shaped Concepts
S = S1.0 ~
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 37
Formalizing Star-Shaped Concepts
S = S1.0 ~ S0.5 ~
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 38
Formalizing Star-Shaped Concepts
S = S1.0 ~ S0.5 S0.25 ~ ~
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 39
Intersection of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 40
Intersection of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 41
Intersection of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 42
Intersection of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 43
Intersection of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 44
Intersection of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 45
Intersection of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 46
Unification of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 47
Unification of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 48
Unification of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 49
Unification of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 50
Unification of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 51
Unification of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 52
Unification of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 53
Unification of Two Concepts
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 54
Projection of a Concept
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 55
Projection of a Concept
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 56
Research Contributions
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 57
Research Contributions
- We can encode correlations in a geometric way
- Most formalizations of conceptual spaces ignore cross-domain
correlations
- [Rickard2006] considers correlations, but not in a geometric way
[Rickard2006] Rickard, J. T. A Concept Geometry for Conceptual Spaces. Fuzzy Optimization and Decision Making, Springer Science + Business Media, 2006, 5, 311-329
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 58
Research Contributions
- We can encode correlations in a geometric way
- Most formalizations of conceptual spaces ignore cross-domain
correlations
- [Rickard2006] considers correlations, but not in a geometric way
- Easily implementable and computationally efficient
- Cuboid can be represented by two support points
- Single constraint: cuboids of a concept must intersect
[Rickard2006] Rickard, J. T. A Concept Geometry for Conceptual Spaces. Fuzzy Optimization and Decision Making, Springer Science + Business Media, 2006, 5, 311-329
Thank you for your attention!
Questions? Comments? Discussions? https://www.lucas-bechberger.de @LucasBechberger
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 60
Example: The Color Domain
https://en.wikipedia.org/wiki/HSL_and_HSV#/media/File:HSL_color_solid_dblcone_chroma_gray.png
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 61
Intersection & Union (Fuzzy Case)
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 62
Example: Fruit Space
round sweet hue round hue sweet
- range
- range
Granny Smith Granny Smith apple apple lemon lemon pear pear banana banana
A Thorough Formalization of Conceptual Spaces / Bechberger and Kühnberger 63