A Cognitive Heuristic model for Local                        Community Recognition                                        ...
A Cognitive Heuristic model of Local Community RecognitionSummary: • The “ambiguous” concept of Community: just some Human...
A Cognitive Heuristic model of Local Community RecognitionThe “ambiguous” concept of Community: just some Human exampleThe...
A Cognitive Heuristic model of Local Community RecognitionThe “ambiguous” concept of Community: just some Human exampleThe...
A Cognitive Heuristic model of Local Community RecognitionThe “ambiguous” concept of Community: the Clustering Spectrum   ...
A Cognitive Heuristic model of Local Community RecognitionThe Human Social Skills: the perfect community recognizerHumans ...
A Cognitive Heuristic model of Local Community RecognitionA new operative framework for the modeling of Human Cognitive He...
A Cognitive Heuristic model of Local Community Recognition The Human Cognitive Heuristics: an operative definition Using th...
A Cognitive Heuristic model of Local Community Recognition “A Cognitive inspired Community Recognition Algorithm”Consideri...
A Cognitive Heuristic model of Local Community RecognitionThe Simple Case                                 CHALLENGE(      ...
A Cognitive Heuristic model of Local Community RecognitionThe Simple Case                                    "COGNITIVE"DE...
A Cognitive Heuristic model of Local Community RecognitionThe Simple Case                                         ALGORITH...
A Cognitive Heuristic model of Local Community RecognitionThe Simple Case            KNOWLEDGE)DISCOVERY)PHASE)       Enco...
A Cognitive Heuristic model of Local Community RecognitionThe Simple Case                  LEARNING(PHASE(           Learn...
A Cognitive Heuristic model of Local Community RecognitionThe Simple Case                                 INFERENCEPHASE  ...
A Cognitive Heuristic model of Local Community RecognitionThe Simple Case                     THE$ENVIRONMENT$            ...
A Cognitive Heuristic model of Local Community RecognitionThe Simple Case                               AWASS 2012        ...
A Cognitive Heuristic model of Local Community RecognitionThe Simple Case                               AWASS 2012        ...
A Cognitive Heuristic model of Local Community RecognitionThe Simple Case                               AWASS 2012        ...
A Cognitive Heuristic model of Local Community RecognitionThe Simple Case                               AWASS 2012        ...
A Cognitive Heuristic model of Local Community RecognitionThe Simple Case                              FUTURE&STEPS&      ...
A Cognitive Heuristic model of Local Community RecognitionA more Complex Case                                Fundamental  ...
A Cognitive Heuristic model of Local Community RecognitionA more Complex Case                                             ...
A Cognitive Heuristic model of Local Community RecognitionA more Complex Case                                             ...
A Cognitive Heuristic model of Local Community RecognitionA more Complex Case                                           Th...
A Cognitive Heuristic model of Local Community RecognitionA more Complex Case                                             ...
A Cognitive Heuristic model of Local Community RecognitionA more Complex Case                                            T...
A Cognitive Heuristic model of Local Community RecognitionA more Complex Case                                             ...
A Cognitive Heuristic model of Local Community Recognition  A more Complex Case                                           ...
A Cognitive Heuristic model of Local Community RecognitionA more Complex Case                          Preliminary Results...
A Cognitive Heuristic model of Local Community RecognitionA more Complex Case                                             ...
A Cognitive Heuristic model of Local Community RecognitionA step forward: Some open problems - Scalability of the algorith...
Upcoming SlideShare
Loading in …5
×

3 a cognitive heuristic model of community recognition final

434 views

Published on

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
434
On SlideShare
0
From Embeds
0
Number of Embeds
3
Actions
Shares
0
Downloads
5
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • \n
  • 3 a cognitive heuristic model of community recognition final

    1. 1. A Cognitive Heuristic model for Local Community Recognition A. Guazzini* Department of Psychology, University of Florence *: CSDC, Centre for the study of Complex Dynamics, University of Florence, ItalyContacts: andrea.guazzini@complexworld.net emanuele.massaro@complexworld.net franco.bagnoli@complexworld.net Webpage: http://www.complexworld.net/
    2. 2. A Cognitive Heuristic model of Local Community RecognitionSummary: • The “ambiguous” concept of Community: just some Human examples • The Cognitive Skills that make us smart and effective community detectors • The Human Cognitive Heuristics: an operative definition • A new operative framework for the modeling of Human Cognitive Heuristics:The tri-partite model • The challenge • A minimal description of a cognitive inspired community recognizer • Numerical simulations: the recipe • Results • A step forward • Some Open Problems .... AWASS 2012 Edinburg 10th-16th June
    3. 3. A Cognitive Heuristic model of Local Community RecognitionThe “ambiguous” concept of Community: just some Human exampleThe concept of Human Community has been definitelyproved to be too wide and multidimensional to be easily bound into a strict operative definition. AWASS 2012 Edinburg 10th-16th June
    4. 4. A Cognitive Heuristic model of Local Community RecognitionThe “ambiguous” concept of Community: just some Human exampleThe concept of Community appears as Culture dependent and determined by many socio demographic factors AWASS 2012 Edinburg 10th-16th June
    5. 5. A Cognitive Heuristic model of Local Community RecognitionThe “ambiguous” concept of Community: the Clustering Spectrum N°of Communities (K Individuals) A better description for the Human communities ⇠K = 2 structure could be obtained considering the Clustering Spectrum ⇠ K = 1 10 ⇠ K = 4 10 Each Human Social Network can be described in terms of density of ⇠ K interactions among its members, so = 8 10 designing a hierarchy of structures. 1 1 Normalized Weight Among Subjects (i.e. probability of interaction) 0 AWASS 2012 Edinburg 10th-16th June
    6. 6. A Cognitive Heuristic model of Local Community RecognitionThe Human Social Skills: the perfect community recognizerHumans have evolved their cognitive systems immersed into an “Highly Social Environment”, developing “Adapted” and sometimes Dedicated Neural Circuitsfor facing with the Social Problems ... at least within the Typical Sizes of the Human Communities. Humans are: 15 5 effective Community Recognizer: usually they are very “confident” about the communities they belong to and very “confident” about the peculiarities that define Dunbar Theory 15 and distinguish such communities. (Categorization) Evolution has produced a cognitive hierarchy of ecological (typical) social structures. effective Community Detectors: once trained cognition Such structures (Circles) can be defined in terms of Emotional 50 appears as able to reveal an existing/known object Closeness among its members (community) in an effective way, e.g. starting from few and revealed analyzing the elements and consuming few time/resources frequencies of contact. 150 AWASS 2012 Edinburg 10th-16th June
    7. 7. A Cognitive Heuristic model of Local Community RecognitionA new operative framework for the modeling of Human Cognitive Heuristics: The tri-partite model Reaction time Module I Flexibility Unconscious knowledge perceptive and attentive processes Cognitive costs Relevance Heuristic Module II Reasoning Goal Heuristic External Recognition Heuristic Solve Heuristic Data Module III Learning Behavior Evaluation Heuristic The minimal structure of a Self Awareness cognitive agent AWASS 2012 Edinburg 10th-16th June
    8. 8. A Cognitive Heuristic model of Local Community Recognition The Human Cognitive Heuristics: an operative definition Using the theoretical tools of the Cognitive Neurosciences, Community Recognition/Definition and Community Detection can be designed as the ability of the cognitive system to extract relevant information from the environment, creating Prototypes (Mental Schemes) of Perceptive/knowledge Information Pattern Prototype of Cognitive HeuristicsWorld Perception Gate Standard Neural Cognitive Prototype Reasoning Network Module (Mental Scheme-A) I1 P1 w1,1 A1 Relevance/Coherence Conscious Processing Assessment I2 P2 w.,2 A2 K1 w2,1 . Neuro . . K2 . Biology w2,n(K) . wn(i),2 . . of wn(a),2 . Encoding . w.,n(a) . Kn(K) . Pn(i) An(a) wn(i),n(a) . . k1 wn(k),n(a) The Mental Scheme are . k2 activated by the inputs and . changes the representation of IN Kn(k) the environment Bounded Knowledge AWASS 2012 Bounded Knowledge that integrates the Edinburg 10th-16th June that represents the Input Input
    9. 9. A Cognitive Heuristic model of Local Community Recognition “A Cognitive inspired Community Recognition Algorithm”Considering an unknown dynamics network of relations, can be designed a Cognitive Agent that throughoutthe “ecological interactions” with its neighbors, autonomously develops a representation/map of the existing communities, or at least of its “position” along a given dimension? 5 15 50 150Such algorithm should be intrinsically local and hence an optimal “Scalable Community Detection Algorithm” AWASS 2012 Edinburg 10th-16th June
    10. 10. A Cognitive Heuristic model of Local Community RecognitionThe Simple Case CHALLENGE( To#develop#an#algorithm#for#community#detec5on,# accordingly#with#the#cogni5ve#theories#of#probabilis5c# reasoning,#characterized#by#the#following#proper5es/ a<ributes:# • To#be#inherently#Local# # • To#be#characterized#by#a#bounded#ra5onality#(Here# # Memory)# • To#be#able#to#merge#both#individual#and#collec5ve# # knowledge#in#order#to#solve#the#task.#### AWASS 2012 Edinburg 10th-16th June
    11. 11. A Cognitive Heuristic model of Local Community RecognitionThe Simple Case "COGNITIVE"DESIGN" A Our$modeling$approach$to$the$cogni2ve$heuris2cs$first$imposes$a$characteriza2on$ of$the$inner$structure$of$the$atomic$elements$(nodes).$ …..$ Memory$ J1" =  nowledge$representa2on$(Bounded$Memory$Vector)$ K …..$ M Heuris2cs$ =  ncoding$(func2on)$ E =  earning$(func2on)$ L J2" = nference$(func2on)$ I M J3" i" M M1" M1" M2" M2" Answer" M3" M3" H3$ H2$ M4" M4" M.." M.." H1$ MB" MB" Inference$Heuris2cs$ Learning$Heuris2cs$ Encoding$Heuris2cs$ AWASS 2012 Edinburg 10th-16th June
    12. 12. A Cognitive Heuristic model of Local Community RecognitionThe Simple Case ALGORITHM* At#each#(me#step#node#plays#a#four#step#procedure:# • Knowledge#discovery#phase# # • Learning#phase#(Memory#Management)# # • Inference#phase# # • Cogni(ve#Dissonance#Evalua(on#Phase# # The#model#we#propose#depend#on#three#main#parameters:# • SM:#Is#the#maximum#size#of#the#node’s#knowledge#vector#(Memory)# # • α:#Is#a#decay#parameter#which#mimics#the#effect#of#the#“social#distance”# # • m:#Is#a#learning#rate#factor#which#rules#the#speed#of#learning### # AWASS 2012 Edinburg 10th-16th June
    13. 13. A Cognitive Heuristic model of Local Community RecognitionThe Simple Case KNOWLEDGE)DISCOVERY)PHASE) Encoding(Heuris.cs( In(our(first(approxima.on(the(node(interact(only(with(its((firsts)(neighbours( weigh.ng(their(influence(by(a(decay(factor((α).( Cij(=(Connec.vity(Matrix( Mi.((=(Memory(vector(for(subject(i( Ki((=(Incoming(knowledge(vector(for(subject(I( α(=(Decay(factor( K i = (M × C) i. ⋅ α AWASS 2012 Edinburg 10th-16th June
    14. 14. A Cognitive Heuristic model of Local Community RecognitionThe Simple Case LEARNING(PHASE( Learning(Heuris,cs((inspired(by(Availability(Heuris,cs)(( The(incoming(knowledge(vector(is(combined(with(the(Memory(Vector(using(a( learning(rate(factor((m):( t +1 t t M i. = M ⋅ m + K ⋅ (1 − m) i. j Bounding(and(Expanding(Phase( The(bounded(memory(is(implemented(by(considering(only(the(greatest(SM( elements(of(the(Memory(Vector.(Following(the(Availability(heuris,cs(is(shaped(by( the(normaliza,on(of(the(Memory(Vector,(which(expands(the(greatest(elements( and(compresses(the(others.( € Bounding(Algorithm:( ( ( ( ( Availability(Heuris,cs((Normaliza,on)( [a(b]=(sort(Mi.,’descend’)( 1 M i,b(S M :length(b )) = 0 M i. = M i. ⋅ N ∑M ij j =1 AWASS 2012€ Edinburg 10th-16th June
    15. 15. A Cognitive Heuristic model of Local Community RecognitionThe Simple Case INFERENCEPHASE InferenceHeuris,cs(inspiredbyRepresenta,venessHeuris,cs) Inferencephasehasadoublerole.Theformeristoproduceaninferenceabout thelocalstructureoftheirnetwork,thela@eristhees,ma,onofthereliabilityof theinferenceitselfbycompu,ngasortofuncertaintyoftheinforma,on (Cogni,veDissonance). Thesimpleruleforthefirsttaskfollowa“TaketheBest”approach.Eachnodes belongtothesameclusterofitsgreatestmemoryelement. Cogni0veDissonance Inordertoes,matethereliabilityoftheirownknowledgeoftheenvironment eachnodecomputesaweighteddiscrepancyamongtheirmemoryvectorand thosecomingfromitsneighbours,asfollows: N ∑M ij −K i j j =1 ΔSi = N AWASS 2012 Edinburg 10th-16th June
    16. 16. A Cognitive Heuristic model of Local Community RecognitionThe Simple Case THE$ENVIRONMENT$ Let$us$start$with$a$very$simple$approxima3on$of$a$“users$network”$temporary$ characterized$by$a$Sta3c,$Symmetric$and$Un@weighted$structure$of$connec3ons.$ 2$ 1$ 4$ 6$ 3$ 5$ 7$ 8$ 10$ 12$ 9$ 11$ AWASS 2012 Edinburg 10th-16th June
    17. 17. A Cognitive Heuristic model of Local Community RecognitionThe Simple Case AWASS 2012 Edinburg 10th-16th June
    18. 18. A Cognitive Heuristic model of Local Community RecognitionThe Simple Case AWASS 2012 Edinburg 10th-16th June
    19. 19. A Cognitive Heuristic model of Local Community RecognitionThe Simple Case AWASS 2012 Edinburg 10th-16th June
    20. 20. A Cognitive Heuristic model of Local Community RecognitionThe Simple Case AWASS 2012 Edinburg 10th-16th June
    21. 21. A Cognitive Heuristic model of Local Community RecognitionThe Simple Case FUTURE&STEPS& • SM!!$!m!–!α!:!might!be!posed!as!dynamic!parameters!used!by!Cogni:ve! ! Heuris:cs!to!explore!efficiently!the!network.! • To!take!into!account!Asymmetric,!Weighted!and!Dynamical!networks.! ! • To!make!the!algorithm!scalable!through!appropriate!Heuris:cs! ! Strategies!based!on!Cogni:ve!DIssonance!!! AWASS 2012 Edinburg 10th-16th June
    22. 22. A Cognitive Heuristic model of Local Community RecognitionA more Complex Case Fundamental Developments - Heterogeneous and dynamics parameters “m” and “alpha”. - Introduction of a Typical Time Scales (e.g. Circadian Rhythm) in correspondance of which the State Vector is reset. - Introduction of a Bounded Long Term Knowledge Vector AWASS 2012 Edinburg 10th-16th June
    23. 23. A Cognitive Heuristic model of Local Community RecognitionA more Complex Case The Agent Random Memory Random Learning Parameter Parameter Short Term (Unconscious) “Bounded” Knowledge mi 2 (0, 1) i 2 (1, 1) Long Term (Conscious) “Bounded” Knowledge S1 K1,1 K1,2 ... K1,n(s) S2 K2,1 . . . . . . . . . . . . Sn(S) Kn(K),1 . . Kn(K),n(s) State Bounded Vector Si(t) Knowledge Bounded Vector Ki(t)where n(s) is a finite constant where n(K) is a finite constant XN Agent Estimated Entropy Agent Cognitive Dissonance N X Ei = t Sij log(Sij ) t t Di,j = t t |Si,k t Sj,k | j=1 k=1 AWASS 2012 Edinburg 10th-16th June
    24. 24. A Cognitive Heuristic model of Local Community RecognitionA more Complex Case The Environment Connectivity Matrix Connectivity Matrix 10 Relevant Features 10 20 N = 90 20 30 30 40 Large Comm (BC)= 1 (90) 40 50 Medium Comm (MC) = 5 (18) 50 60 60 70 Small Comm (SC) = 10 (9) 70 80 80 90 P(Lij)=PA with PA(BC)< PA(MC)< PA(SC) 90 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 Unweighted Network (Adjacency Matrix) Three different “Typical Sizes” AWASS 2012 Edinburg 10th-16th June
    25. 25. A Cognitive Heuristic model of Local Community RecognitionA more Complex Case The Recipe 1- Discovery Phase Information Spreading/Gathering phase and State Vector (Unconscious Knowledge) Updating 2- Cognitive Dissonance Phase Evaluation of Ego-side Information Entropy and Cognitive Dissonance (with neighbors) 3- Reasoning Phase Evolution/Modification of the parameters whenever the discovery phase is “mute” 4- Inference Phase Synchronized Reset of all the State Vector and Extrapolation of the first K relevant “approximation” of the network (state vectors), by the exploitation of the Ego Entropy. AWASS 2012 Edinburg 10th-16th June
    26. 26. A Cognitive Heuristic model of Local Community RecognitionA more Complex Case The Recipe 1- Discovery Phase Information Spreading/Gathering phase and State Vector (Unconscious Knowledge) Updating Gathering SubPhase Learning SubPhase t (Qi,j ) i t t+1 Si,j = Pk(i) t k=1 (Qi,k ) t i k(i) X Qt = mt Si + (1 i i t mt ) i t Sk k=1 Expansion of biggest component and reduction of smallest component by renormalization. Where S is the state vector, k(i) is the number of neighbors of the agent i, and mti the memory of agent i at time t AWASS 2012 Edinburg 10th-16th June
    27. 27. A Cognitive Heuristic model of Local Community RecognitionA more Complex Case The Recipe 2- Cognitive Dissonance Phase Evaluation of Ego-side Information Entropy and Cognitive Dissonance (with neighbors) XN Agent Estimated Entropy Agent Cognitive Dissonance N X t Ei = Sij log(Sij ) t t Di,j = t t |Si,k t Sj,k | j=1 k=1 AWASS 2012 Edinburg 10th-16th June
    28. 28. A Cognitive Heuristic model of Local Community RecognitionA more Complex Case The Recipe3- Reasoning PhaseEvolution/Modification of the parameters whenever the discovery phase is “mute” and detection of thechange of sign in the second derivative of the entropy (Eti) IF Just a “stupid/smart” rule Pk(i) Pk(i) ds=0.1; T X t 1 t k=1 Di,k Then k=1 Di,k m(1,i)=m(1,i)*abs((randn*ds)+1); t |(Ei 1 + )| |(Ei + t )| < if m(1,i)>1, m(1,i)=1; end; t=t⇤ k(i) k(i) if m(1,i)<0, m(1,i)=0.01; end; FOR T t⇤ > t⇤ alpha(1,i) = 1.5*abs((randn*ds)+1); alpha(1,alpha(1,i)<1)=1; When the sign of the second derivative of the Agent Entropy changes, the node temporary registers respectively: - The state Vector - The value of the first derivative of Entropy - The absolute Value of the Entropy - The Cognitive Dissonance Time AWASS 2012 Edinburg 10th-16th June
    29. 29. A Cognitive Heuristic model of Local Community Recognition A more Complex Case The Recipe 4- Inference Phase Synchronized Reset of all the State Vectors and Extrapolation of the firsts K relevant “approximation” of the network (state vectors), by the exploitation of the Ego Entropy. Sample coming from a “typical” discovery period (in humans the day)Knowledge Time Bounded Rationality AWASS 2012 Edinburg 10th-16th June
    30. 30. A Cognitive Heuristic model of Local Community RecognitionA more Complex Case Preliminary Results AWASS 2012 Edinburg 10th-16th June
    31. 31. A Cognitive Heuristic model of Local Community RecognitionA more Complex Case Subject i (i=3) Long Preliminary Results Term bounded Memory AWASS 2012 Edinburg 10th-16th June
    32. 32. A Cognitive Heuristic model of Local Community RecognitionA step forward: Some open problems - Scalability of the algorithm with the System Size (N) - Validation of the Dunbar Theory about the existence of typical sizes of the human communities, due to their cognitive limits (i.e. Bounded Rationality) and the environmental constraints (i.e. Network Topology) - Multidimensional (i.e. more ecological) State Vector - Rewiring, Pruning and human heuristics for the Network Management. AWASS 2012 Edinburg 10th-16th June

    ×