1 / 15

Priscila M. V. Lima , Felipe M. G. França and Guilherme M. B. Domingues UFRJ

Priscila M. V. Lima , Felipe M. G. França and Guilherme M. B. Domingues UFRJ 1 st EELA Grid School Itacuruç á , June 4 th , 2006. G-SATyrus: A Gridfied SAT-based Neuro-Symbolic Problem Solving Architecture Task 3.3: Additional Applications. SATyrus. SAT yrus:

eshe
Download Presentation

Priscila M. V. Lima , Felipe M. G. França and Guilherme M. B. Domingues UFRJ

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Priscila M. V. Lima, Felipe M. G. França and Guilherme M. B. Domingues UFRJ 1st EELA Grid School Itacuruçá, June 4th, 2006 G-SATyrus: A Gridfied SAT-based Neuro-Symbolic Problem Solving ArchitectureTask 3.3: Additional Applications

  2. Itacuruçá, June 25th, 2006 SATyrus SATyrus: • SATisfiability-based, neuro-symbolic architecture; • Exact formulation synthesizer. Target Problem, e.g., colouring Constraints Modeling SATyrus compiler Energy Function Optimizer, e.g., stochastic HONNs Global/Local optima

  3. Itacuruçá, June 25th, 2006 SATyrus Target Problem, e.g., colouring Graph Colouring consists on determining the minimum assignment of colours to the vertices of G=(V, A); G=(V, A)

  4. Itacuruçá, June 25th, 2006 SATyrus Target Problem, e.g., colouring Constraints Modeling Integrity Constraints: • (i) Every vertex must have one colour assigned to it: • ∀i, ∀k| 1 ≤ i ≤ n, 1 ≤ k ≤ n: ∨(vcik). • So, letφ5 = ∧i(∨k (vcik)). • (ii) Two neighbouring vertices cannot have the same colour: • ∀i, ∀i’, ∀k| 1 ≤ i ≤ n, 1 ≤ i’ ≤ n, 1 ≤ k ≤ n, i ≠ i’: ¬(neighii’) ∨ ¬(vcik ∧vci’k). • So, letφ6 = ∧i∧i’≠i∧k(¬(neighii’) ∨ ¬(vcik ∧vci’k)).

  5. Itacuruçá, June 25th, 2006 SATyrus Target Problem, e.g., colouring Constraints Modeling Integrity Constraints (cont.): • (iii) A vertex cannot have more than one colour: • ∀i, ∀k, ∀k’ | 1 ≤ i ≤ n, 1 ≤ k ≤ n, 1 ≤ k’ ≤ n, k ≠ k’: ¬(vcik∧ vcik’). • So, letφ7 = ∧i∧k∧k’≠k ¬(vcik∧ vcik’). • (iv) If a colourkis assigned to a vertex in matrixVcolour, then the corresponding unit in matrixCcolourmust be activated: • ∀i, ∀k | 1 ≤ i ≤ n, 1 ≤ k ≤ n: ¬vcik∨ ck. • So, letφ8 = ∧i∧k(¬vcik∨ ck).

  6. Itacuruçá, June 25th, 2006 SATyrus Target Problem, e.g., colouring Constraints Modeling Optimality Constraint: • (v) The number of activated elements in matrix Ccolour: • ∀k | 1 ≤ k ≤ n : ck. • So, let φ9 = ∨k ck.

  7. Itacuruçá, June 25th, 2006 SATyrus Target Problem, e.g., colouring Constraints Modeling Constraints’ penalties: • Constraints (iii):  • Constraints (i), (ii) and (iv):  • Constraints (v): 1

  8. Itacuruçá, June 25th, 2006 SATyrus Target Problem, e.g., colouring Constraints Modeling %GRAPH COULORING(SATyrus specification) num=6; neigh(num,num); vc(num,num); colour(num); integrity group type alfa: Forall {i,k}; 1 <= I <= num,1 <= k <= num: vc[i][k]; integrity group type beta: Forall {i,l,k}; 1 <= I <= num,1 <= l <= num,1 <= k <= num; I != l: (not neigh[i][l] or not vc[i][k]or not vc[l][k]); integrity group type beta: Forall {i,k,m}; 1 <= I <= num,1 <= k <= num;1$<=$m$<=$num;k$!=$m: (not vc[i][k] or not vc[i][m]); integrity group type beta: Forall {i,k}; 1 <= I <= num,1 <= k <= num: (not vc[i][k] or colour[k]); optimality group type costo: Forall {k}; 1 <= k <= num: colour[k]; PENALTY { beta is level 2; alfa is level 1; costo is level 0;} SATyrus compiler

  9. Itacuruçá, June 25th, 2006 SATyrus Target Problem, e.g., colouring Constraints Modeling SATyrus compiler Energy Function

  10. Itacuruçá, June 25th, 2006 SATyrus Motivations: Hard, composite, complex problems • Energy Generation Expansion LOA — Advanced Optimization Lab, CNPq, Brazil, 2006; E.g., unification of short, medium and long term models. • Airport Management (surface) CAPES-COFECUB, UFRJ — ENAC&UT2, Toulouse, 2006 (submitted); E.g., A-SMGCS support: routing, monitoring, signaling and control of mobile entities.

  11. Itacuruçá, June 25th, 2006 SATyrus • Artificial Logical Reasoning ARQ-PROP2, ARQ-FOL — Resolution based, ANN based architectures (Lima, 1992) (Lima, 2000) (Lima, 2001a) (Lima, 2001b) ; • 3D Molecular Structure Reconstruction Molecular reconstruction via integration of independent models; Equivalence between geometrical model and classic molecular model proven (Glaucia Pereira, 2006);

  12. Itacuruçá, June 25th, 2006 SATyrus • Grid OS Dynamic and Distributed Scheduling of Execution, Storage and Communication Resources; Distributed File/Storage System; Intelligent/Reactive Communication System; (Diego Carvalho, Guilherme Domingues and Pedro Henrique Rausch Bello);

  13. Itacuruçá, June 25th, 2006 SATyrus First grid enabling strategy: Multistart Single HONN under slow cooling; geometrical cooling (0.99) versus Multiple HONNs starting from different initial points; cool runnings ;-) • Adv.: copes with lack of agile inter-domain communication capabilities (bag-of-tasks); • Disadv.: problems limited by single CPU capabilities.

  14. Itacuruçá, June 25th, 2006 SATyrus Second grid enabling strategy (perhaps EELA-2): Fully Distributed Single HONN under slow cooling spread throughout the grid; • Adv.: interesting (and huge)problems, e.g., 3D molecular reconstruction (110 PetaBytes), could be tackled; not limited by single CPU capabilities. • Disadv.: does not cope with lack of agileinter-domain communication capabilities;

  15. Itacuruçá, June 25th, 2006 SATyrus Thanks. priscila@nce.ufrj.br felipe@cos.ufrj.br

More Related