420 likes | 496 Views
Global Networks in Computer Science?. Guerino Mazzola U & ETH Zürich guerino@mazzola.ch www.encyclospace.org . Motivation Local Networks Global Networks Diagram Logic. Motivation Local Networks Global Networks Diagram Logic.
E N D
Global Networks in Computer Science? Guerino Mazzola U & ETH Zürich guerino@mazzola.ch www.encyclospace.org
Motivation • Local Networks • Global Networks • Diagram Logic
Motivation • Local Networks • Global Networks • Diagram Logic
Course by Harald Gall: Soft-Summer-Seminar 31.8./1.9. 2004 SW-Architekturen/Evolution „Klassifikation von Netzwerken...“
Transformational Theory, K-nets (Lewin et al.) sets of notes Perspectives of New Music (2006) Guerino Mazzola & Moreno Andreatta: From a Categorical Point of View: K-nets as Limit Denotators
— U T — torus T compact Open set U not compact T U manifolds = global objects in differential geometry
Motivation • Local Networks • Global Networks • Diagram Logic
h h‘ = A V E = B W d v t t‘ w u q x = t(a) c a b x y = h(a) a y Digraph = category of digraphs (= quivers, diagram schemes, etc.) Digraph(,E)
Dllr Di allr i Dilq ailq Dl l Dlip alip Dijt aijt D Djlk ajlk j Dj Djms ajms C m Dm Diagram in a category C = digraph morphismD: C • Di = objects in C • Dijt = morphisms in C
Examples: • diagram of sets C =Set • diagram of topological spaces C =Top • diagram of real vector spaces C =Lin— • diagram of automata C = Automata • etc.
C@ @ @C • Yoneda embedding • Let C@ = category of contravariant functors (= presheaves) F: C Set • Have Yoneda embedding functor @:C C@ • @X: C Set: A ~> A@X = C(A, X) (@X = representable presheaf) C
F x A h F G A B x y address change • Category ∫C of C-addressed points • Objects of ∫C • x: @A F, F = presheaf in C@~xF(A), write x: A F A = address, F = space of x • Morphisms of ∫C • x: A F, y: B G h/: x y
hllr/llr hllr/llr xi: Ai Fi xi: A F hilq/ilq hilq/ilq xl: Al Fl xl: AF hlip/lip hlip/lip hijt/ijt hijt/ijt hijt/ijt hjlk/jlk hjlk/jlk D xj: Aj Fj xj: AF C@ hjms/jms hjms/jms xm: Am Fm xm: AF coordinateofx Local network in C= diagram x of C-addressed points x: ∫C xlim(D) x is flat if all addresses and spaces coincide.
Ÿ12 Ÿ12 Ÿ12 2 3 T4/Id 7 3 Ÿ12 Ÿ12 0 0 4 7 T11.5/Id T11.-1/Id 0 0 2 4 T2/Id Example 1: K-nets of pitch classesC = Ab abelian groups + affine maps
2Ÿ12 2Ÿ12 2T4/Id {2,7,8} {3,4,10} 0 0 2T11.5/Id 2T11.-1/Id 0 0 {1,2,7} {3,4,9} 2T2/Id 2Ÿ12 2Ÿ12 Example 2: K-nets of chordsC = Ab
Ÿ12 Ÿ12 Id/T11.-1 Ks s Ÿ12 Ÿ12 Ÿ11 Ÿ11 T11.-1/Id T11.-1/Id Ÿ11 Ÿ11 s Us UKs Id/T11.-1 Example 3: K-nets of dodecaphonic seriesC = Ab
2004 Example 4: Neural Networks
h —n —m x y Ÿ Ÿ +? Neural Networks C = Set address = Ÿ Points x: Ÿ —nat this address are time series x = (x(t))t of vectors in —n.They describe input and output for neural networks. Dn = Ÿ @ —n h/+? : x y y(t) = h(x(t-1))
D h Dn Dn D Dn p1 p1 p12 p3 Dn Dn Dn Dn D D D D Dn pi Id/+? Id/+? p2 o a ?,? pn D
x1 h p1 p1 p12 p3 xi pi Id/+? Id/+? ?,? p2 a o pn xn (+w,+x, a+w,+x) w o(a+w,+x) (w, x) +w,+x a+w,+x x (+w,+x)
2004 (e) = f 2 Example 5: Local Networks of Automata • C = AutomataSet S of states, alphabet A • Objects: (e, M: S A 2S) • Morphisms: h = (, ): (e, M: S A 2S) (f, N: T B 2T) S A 2S T B 2T
hllr/Id si: A Mi hilq/Id sl: A Ml hlip/Id hijt/Id hjlk/Id sj: A Mj hjms/Id sm: A Mm addressA = (0, M: {0,1} 2{0,1}) points x: A (e, M: S A 2S) ~ states s in S local network ofA-addressed pointsIdA = address change~ network of states
Class@ virtual classes objective classes Example 6: Networks of OO Instances • C = Class classes and instances of a OO language • Objects: classes and one special address: I = „the instance“ (corresponds to final object 1) • Morphisms: s: K L superclass v: K F field m: K M method (without arguments) i: I K instance • I@K = {instances of class K } @Class
m Id @M F m(i,j) pK pL @K @L (i,j) I I i j Cartesian product multiple inheritance Instance method in two variables: F = @K @L (i,j):I F, m: F @M
xllr xi xilq xl xlip xijt di x = xjlk xj yf(i) yrrh yf(i)rq xjms yrf(i) p yr xm yf(i)st ysrw ys y= Morphisms of local networks x: ∫C, y: E ∫C f: x y category LC f: E for every vertex i of , there is a morphism di: xi yf(i) subcategory FC Flat morphism: x, y flat and di = const. = h/
Special cases • identity morphism Idx: x x • isomorphisms f: x y there is g: y x with g∞f = Idxund f∞g = Idy, write x y. • local subnetworksLocal network y: E ∫C , f: E subdigraph, f: y y embedding morphism.
Motivation • Local Networks • Global Networks • Diagram Logic
s r rs atlas
chart yi i i l yl l r j yj j ym m m cartes. rs xi i yi l xl yl xj j s yj i isomorphism of local networks l xi xl j xj chart
Examples • Local networks are global networks with one chart. • Interpretations: let y: E ∫C be a local network and letI = (i) be a covering by subdigraphs i E. Build the corresponding subnetworks xi = y i. Together with the identity on the chart overlaps, this defines a global network yI, called interpretation of y.Interpretations are interesting for the classification of networks by coverings of a given type of charts!Visualization via the nerve of the covering. • Locally flat global networks have flat charts and local glueing data.
Morphisms of global networksx, y over category C f: x y= morphisms of their digraphs, which induce morphisms of local networks. • Category GC of global networks over C. • SubcategoryLfCof locally flat networks + locally flat morphisms. • A global networkis interpretable, if it is isomorphic to an interpretation. Open problem: Under what condition are therenon-interpretable global networks? LfC X GC
4 6 x |x| ~> 4 3 6 5 2 3 5 1 2 1 Theorem Given address A in C, we have a verification functor |?|: ALfCredAGlob Corollary There are non-interpretable global networks in ALfCred COLLOQUIUM ON MATHEMATHICAL MUSIC THEORY H. Fripertinger, L. Reich (Eds.) Grazer Math. Ber., ISSN 1016–7692 Bericht Nr. 347 (2005), Guerino Mazzola: Local and Global Limit Denotators and the Classification of Global Compositions
|x| 4 6 4 1* 5 6 2 4 c d 3 b a 5 2 1 3 2* 6 5 3 1
Dendritic transformations Karl Pribram
Motivation • Local Networks • Global Networks • Diagram Logic
1 = Alexander Grothendieck The category Digraph is a topos D E D+E DE 0 = Ø
= v w x y T In particular:The set Sub(D) of subdigraphsof a digraph Dis a Heyting algebra: have „digraph logic“. Ergo: Global networks, ANNs, Klumpenhouwer-nets, and local/global gestures, enable logicaloperators (, , ,) Subobject classifier
Heyting logic on set Sub(y) of subnetworks of y h, k Sub(y)h k := h kh k := h kh k (complicated) h := h Ø tertium datur: h ≤ h u: y1 y2Sub(u): Sub(y2) Sub(y1) homomorphism of Heyting algebras = contravariant functor Sub: LC Heyting Sub: GC Heyting complexes
c b IV V d a III II VI VII I e g f C-major network of degrees y =3.x + 7
V IV I VI I =
Describe global ANNs. • Can we interpret the dendritic transformations in the theory of Karl Pribram as being glueing operations of charts for global ANNs? • What is the gain in the construction of global ANNs? Is there any proper „global“ thinking in such a model? • What can be described in OO architectures by global networks, that local networks cannot? • Was would global SW-engineering/programming mean? How global are VM architectures?