セクションアウトライン


    • (original file : Animals.xdsl)
    • BN used during the course : Animals2.xdsl (generated by replacing logical dependencies by CPDs)

    • Scenario 1 : without evidence
      • initialization of λ vectors : ?
      • first λ messages : λWa(Cl)= ? λBo(Cl)= ? λHa(Cl)= ?     but also λEn(An)= ? λBe(An)= ?
      • aggregation of some λ messages : λ(Cl)=?
      • last λ messages : λCl(An)=?
      • last aggregation : λ(An)=?
      • initialization of π vectors ?
      • first π messages : πEn(An)=? πCl(An)=? πBe(An)=?
      • first π aggregations : π(En)=? π(Cl)=? π(Be)=?
      • other π messages : πWa(Cl)=? πBo(Cl)=? πHa(Cl)=?
      • last π aggregations : π(Wa)=? π(Bo)=? π(Ha)=?
      • and finally, what is the probability of each variable given the (empty) evidence ? (check with Genie)

    • Scenario 2 : with evidence {Class=Mammal, Bears Young As=Eggs}
      • initialization of λ vectors : ?
      • first λ messages :
        • λWa(Cl), λBo(Cl), λHa(Cl) are skipped because Cl is observed
        • λEn(An)= ? λBe(An)= ? λCl(An)= ?
      • aggregation : λ(An)=?
      • initialization of π vectors ?
      • first π messages :
        • πCl(An) and πBe(An) are skipped because Cl is observed
        • πEn(An)=?
      • first π aggregation : π(En)
      • other π messages : πWa(Cl)=? πBo(Cl)=? πHa(Cl)=? you can notice that the information about Be is not usefull here !
      • last π aggregations : π(Wa)=? π(Bo)=? π(Ha)=?
      • and finally, what is the probability of each variable given the (empty) evidence ? (check with Genie)

    • Scenario 3 : with evidence {Warm Blooded=true, Bears Young As=Eggs}
      • do it by yourself :-)