![Page 1: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/1.jpg)
Learning bounded unions of Noetherian closed set systems
via characteristic sets
Yuichi Kameda1, Hiroo Tokunaga1 and Akihiro Yamamoto2
1Tokyo Metropolitan University2Kyoto University
![Page 2: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/2.jpg)
Key Concepts In this talk “learning” is identification in the li
mit from positive data. Noetherian Closed Set Systems are classes
of languages with algebraic structure as targets of learning.
Characteristic sets are used for sufficient conditions for learnability of language classes. They have also another meaning in regarding languages as algebraic objects.
![Page 3: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/3.jpg)
Results Previous work:
It is shown that bounded unions of languages are learnable theoretically[Kobayashi, Kapur et al.] and several learning procedures have been proposed on various viewpoints.
This talk: We give a schema of learning procedure on bounded unions of certain class of languages, called Noetherian closed set system, by using characteristic sets.
![Page 4: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/4.jpg)
Outline Inductive Inference and Algebra Precise Definitions of Noetherian Closed S
et System and Characteristic sets Main Result Example
![Page 5: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/5.jpg)
The purpose of our research Clarifying the properties of
learning languages with algebraic structure on both the viewpoint of learning and mathematics.
Current subjectLearning bounded unions of languages, which have not attracted mathematicians, but are very popular for learning people.
![Page 6: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/6.jpg)
Outline Inductive Inference and Algebra Precise Definitions of Noetherian Closed S
et System and Characteristic sets Main Result Example
![Page 7: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/7.jpg)
Inductive Inference and Algebra Some learning people have investigated learni
ng languages with algebraic structure: ideals of Z [Angluin] pattern languages [Angluin] ideals in commutative rings [Stephen & Vents
ov]
![Page 8: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/8.jpg)
Examples72, 48, 60, …, 12,…
Hypotheses 72, 24, 12,…
72
Teacher
Learning Machine
Identification of Ideals of Z from positive data
Class of Ideals
・・・
n
n= { nx | x is an integer }
Computing GCD of Given Examples
![Page 9: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/9.jpg)
Algebra and Inductive Inference Recently some researchers of History of
Mathematics have found that the original version of Hilbert’s basis theorem can be regarded as learning.This means Mathematician used "learning" in algebra in late 19th century.
In fact, Hilbert’s basis theorem can be regarded as learning.
![Page 10: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/10.jpg)
Hilbert’s original paper
![Page 11: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/11.jpg)
Hilbert’s original paper
![Page 12: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/12.jpg)
Why Set-Union of Ideals? In Mathematics :
The set-union of two ideals has not been interested because it is difficult to give its “meaning”.
In Learning Theory :Finite elasticity of a class of languages ensures identifiability of set union of two languages in it, but does not effective procedure to compute its finite tell-tale.
![Page 13: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/13.jpg)
Outline Inductive Inference and Algebra Precise Definitions of Noetherian Closed S
et System and Characteristic sets Main Result Example
![Page 14: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/14.jpg)
Ideals in Commutative Ring
A subset I of a ring R is an ideal if the followings are satisfied:
0 IIf f I and g I, f + g IIf f I and h R, h f I
Examples n= { xn | x Z } for every n in Z f, g= { h f + kg | h, k Q[X1,..., Xn ]}
for every f, g Q[X1,..., Xn ]
![Page 15: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/15.jpg)
Closed Set Systems(CSS) A mapping C: 2U 2U is a closure
operator if it satisfies:X C(X)C(C(X)) = C(X)X Y C(X) C(Y) (X,Y U).
XU is closed if C(X) X. A closed set system is a class of
closed sets.
![Page 16: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/16.jpg)
Characteristic Sets A finite subset SLL is a characteristic
set of L in L L’L, SL’ LL’.
S L
![Page 17: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/17.jpg)
Finite Elasticity(F.E.) L has infinite elasticity if there exist an infinite sequ
ence of elements s0, s1, … and languages L1, L2, … such that {s0,s1,…,sn-1}Ln and snLn..
L has finite elasticity L does not have infinite elasticity.
s0 s1 s2 sn-1…L1
L2
Ln
L3
![Page 18: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/18.jpg)
Closed Set System and F.E. If L is CSS, thenTheorem[de Brecht et al.]
L has F.E. L satisfies the ascending chain condition, i.e. L has no infinite chain of languages
L1 L⊊ 2 ⊊ … ⊊ Ln …⊊ . A Noetherian closed set system(NCSS) i
s a CSS that has F.E..
![Page 19: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/19.jpg)
Bounded Unions of Languages
L: a class of language.
⋃kL L1⋃… L⋃ mLiL, mk .
Theorem[Motoki-Shinohara-Wright] If L has finite elasticity, then ⋃kL also has.
![Page 20: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/20.jpg)
Bounded Unions of NCSS
L: Noetherian closed set system.
1. ⋃kL has finite elasticity.
2. Every element of ⋃kL has a characteristic set.
3. ⋃kL is identifiable from positive data.
![Page 21: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/21.jpg)
Outline Inductive Inference and Algebra Precise Definitions of Noetherian Closed S
et System and Characteristic sets Main Result Example
![Page 22: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/22.jpg)
Main Result Suppose that L is NCSS and ⋃kL is co
mpact. We give an algorithm schema for learning ⋃kL under the condition:
for LL, a char. set of L in ⋃kLis computable from char. set of L in L.
Compactness
⋃kL is compact iff LL1⋃… L⋃ m i s.t. LLi (LL, L1⋃… L⋃ m⋃kL).
![Page 23: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/23.jpg)
Learning SchemaTarget: L1⋃… L⋃ m⋃kL
Positive data: f1, f2,…, fn,…
Step n: By using hypergraph, construct a hypothesis H⋃kL s.t.
H L1⋃… L⋃ m, H contains elements of {f1,…,fn} as many
as possiblefrom {f1,…,fn}.
![Page 24: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/24.jpg)
Construction of Hypergraph Step 1
Set of examples: {f1}
Set of hyperedges HE1: {{f1}}Set of vertices V1: {f1}
f1
![Page 25: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/25.jpg)
Construction of Hypergraph Step n
Set of vertices Vn: {f1,…,fn} = Vn-1 ⋃ {fn}
Set of examples: {f1,…,fn}
fn
f1
f2
f3
f4
fn-1
1. Set HEn = HEn-1.
![Page 26: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/26.jpg)
Construction of Hypergraph2. For SVn do
If a char. set of C(S) in ⋃kL is contained by Vn, thenAdd S to HEn, and remove all hyperedges contained by S.
3. If no hyperedge has fn, then add {fn}.
fn
f1
f2
f3
f4
fn-1
![Page 27: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/27.jpg)
Learning Algorithm of ⋃kLRepeat1. Put n=1.2. Construct a hypergraph Gn from {f1,…,fn}.
3. Choose at most k maximal hyperedges of Gn w.r.t. some ordering.
4. Output (at-most) k-tuple in 3.5. Add 1 to n.forever.
The ordering at 3 can be taken freely provided that
C(S) C(S’) S S’.
![Page 28: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/28.jpg)
Outline Inductive Inference and Algebra Precise Definitions of Noetherian Closed S
et System and Characteristic sets Main Result Example
![Page 29: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/29.jpg)
Example (Polynomial ideal)Let I be the class of all ideals of polynomial ring Q[x,y].
Target: x2,y3⋃x3,y2⋃2I Positive data:
x2, y3, y2, x2+y3, x3, x3+y2,… For f,gQ[x,y], {f, g, f+g} is a characteri
stic set of f,g in ⋃2I.
![Page 30: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/30.jpg)
Example (Polynomial ideal)
Step 1
Target: x2,y3⋃x3,y2⋃2IPositive data: x2, y3, y2, x2+y3, x3, x3+y2,…
Hypergraph G1:
x2
Output: x2
The set of examples: {x2}
![Page 31: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/31.jpg)
Step 2
Hypergraph G2:
Example (Polynomial ideal)Target: x2,y3⋃x3,y2⋃2IPositive data: x2, y3, y2, x2+y3, x3, x3+y2,…
x2
Output: x2 ⋃ y3
y3
Output: x2
The set of examples: {x2, y3}
![Page 32: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/32.jpg)
Output: x2 ⋃ y3
Hypergraph G3:
Example (Polynomial ideal)
Step 3
Target: x2,y3⋃x3,y2⋃2IPositive data: x2, y3, y2, x2+y3, x3, x3+y2,…
x2
Output: x2 ⋃ y2
y3
y2
(y3 is not maximal: y3 y2sincey3 y2 )
The set of examples: {x2, y3, y2}
C(S) C(S’) S S’.
![Page 33: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/33.jpg)
Output: x2 ⋃ y2
Example (Polynomial ideal)
Step 4
Target: x2,y3⋃x3,y2⋃2IPositive data: x2, y3, y2, x2+y3, x3, x3+y2,…
x2
Output: x2,y3 ⋃ y2
y3
y2
x2+y3
Hypergraph G4:
The set of examples: {x2, y3, y2, x2+y3}
{x2,y3,x2+y3} is a char. set of x2,y3 in ⋃2I.
![Page 34: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/34.jpg)
Example (Polynomial ideal)
Step 5
Target: x2,y3⋃x3,y2⋃2IPositive data: x2, y3, y2, x2+y3, x3, x3+y2,…
x2
Output: x2,y3 ⋃ y2
y3
y2
x2+y3
x3Hypergraph G5:
(x3 is not maximal: x3 x2,y3sincex3 x2,y3 )
The set of examples: {x2, y3, y2, x2+y3, x3}
![Page 35: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/35.jpg)
Output: x2,y3 ⋃ y2
Example (Polynomial ideal)
Step 6
Target: x2,y3⋃x3,y2⋃2IPositive data: x2, y3, y2, x2+y3, x3, x3+y2,…
x2
Output: x2,y3 ⋃ x3,y2
y3
y2
x2+y3
x3
x3+y2
Target language
Hypergraph G6:
The set of examples: {x2, y3, y2, x2+y3, x3, x3+y2}
{y2,x3,x3+y2} is a char. set of x3,y2 in ⋃2I.
![Page 36: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/36.jpg)
Application The algorithm schema can be
applied to learning bounded unions of Tree Pattern Languages.
![Page 37: Learning bounded unions of Noetherian closed set systems via characteristic sets](https://reader036.vdocuments.us/reader036/viewer/2022070502/56813727550346895d9eb235/html5/thumbnails/37.jpg)
Conclusion
We proposed a learning algorithm schema by combining hypergraph and characteristic sets for bounded unions of Noetherian closed set systems.