Causal Bayesian Networks

Loading...

Flash Player 9 (or above) is needed to view presentations.
We have detected that you do not have it on your computer. To install it, go here.

0 comments

Post a comment

    Post a comment
    Embed Video
    Edit your comment Cancel

    Favorites, Groups & Events

    Causal Bayesian Networks - Presentation Transcript

    1. Causal Bayesian Networks Fl´vio Code¸o a c Coelho Basic graph theory Causal Bayesian Networks Bayesian Networks Inference Fl´vio Code¸o Coelho a c Oswaldo Cruz Foundation November 1, 2006
    2. Graphs Causal Bayesian Sets of elements called vertices, V , that may or may not be Networks connected to other vertices in the same set by a set of edges,E Fl´vio Code¸o a c Coelho A graph may be defined uniquely by its set of edges, wich imply Basic graph the set of vertices, e.g.E = {W , X , Y , Z }: theory Bayesian Networks G : E = {(W , Z ), (Z , Y ), (Y , X ), (X , Z )} Inference 1 by running the above code, you’ll get the following output: [ Y , X , Z , W ][( Y , X ), ( Y , Z ), ( X , Z ), ( Z , W )] 1 https://networkx.lanl.gov/
    3. Some properties of graphs Causal Bayesian Networks Fl´vio Code¸o a c Coelho Graphs can be directed or undirected; Basic graph The order of a graph corresponds to its number of vertices; theory Bayesian The size of a graph corresponds to its number of edges; Networks Vertices connected by an edge are neighbors or adjacent; Inference The order of a vertex corresponds to its number of neighbors; A path is a list of edges connecting two vertices; A cycle is a path starting and ending in the same vertex; A graph with no cycles is termed acyclic.
    4. Visualizing the graph Causal Bayesian Networks Fl´vio Code¸o a c Coelho Basic graph theory Bayesian Networks From the code above we get the following picture: Inference
    5. Directed Acyclic Graph (DAG) Causal Bayesian Networks In directed acyclic graphs we use arrows to represent edges. Fl´vio Code¸o a c Coelho Basic graph theory Bayesian Networks The output: Inference
    6. DAG properties Causal Bayesian Networks Fl´vio Code¸o a c Coelho Basic graph Parents,children,descendants,ancestors, etc. theory Root node Bayesian Networks sink node Inference Every DAG has at least one root and one sink Tree graph: every node has at most one parent Chain graph: every node has at most on child Complete graph: All possible edges exist.
    7. Bayesian Networks Causal Bayesian Networks Fl´vio Code¸o a Coelho c Advantages : Basic graph theory 1 Convenient means of expressing assumptions Bayesian Networks 2 economical representation of Joint probabilit functions Inference 3 Facilitate efficient inferences from observations Why Bayesian? 1 Subjective nature of input information 2 Reliance on Bayes conditioning for updating information 3 The distinction between causal and evidential reasoning
    8. Definitions Causal Bayesian Networks Fl´vio Code¸o a c Coelho Markovian parents (PAj ) P(xj | paj ) = P(xj | x1 , . . . , xj−1 ) Basic graph theory such that no subset of PAj satisfies the above Bayesian equation. Networks Inference Markov compatibility If a probability function admits the factorization P(xi | x1 , . . . , xn ) = P(xi | pai ) relative to a DAG G we say that G and P are compatible or that P is Markov relative to G . d-separation Z d-separates X and Y iff Z blocks every path from a node in X to a node in Y .
    9. Theorems Causal Bayesian Networks Fl´vio Code¸o a c Coelho Probabilistic Implications of d-separation If sets X and Y are d-separated by Z in a DAG G , then X is Basic graph theory independent of Y conditional on Z every distribution Bayesian compatible with G . Conversely, if X and Y are not d-separated Networks by Z in a DAG G , then X and Y are dependent conditional on Inference Z in at least one dist. compatible with G . Ordered Markov Condition A necessary and sufficient condition for a probability distribution P to be markovian relative a DAG G is that every variable be independent of all its predecessors in some oredering of the variables that agrees with the arrows of G .
    10. Theorems, cont. Causal Bayesian Networks Fl´vio Code¸o a c Coelho Parental Markov Condition Basic graph A necessary and sufficient condition for a probability theory distribution P to be markovian relative a DAG G is that every Bayesian Networks variable be independent of all its nondescendants (in G ), Inference conditional on its parents. Observational Equivalence Two DAGs are observationally equivalent if and only if they have the same skeletons and the same sets of v-structures, that is, two converging arrows whose tails are not connected by an arrow.
    11. Inference with Bayesian Networks Causal Bayesian Networks Fl´vio Code¸o a c Coelho Basic graph theory In the presence of a set of observations X the posterior Bayesian probability: s P(y , x, s) Networks Inference P(y | x) = y ,s P(y , x, s) can be calculated from a DAG G and the conditional probabilities P(xi | pai ) defined on the families of G
    12. Causal Bayesian Networks Causal Bayesian Networks DAGs constructed around Causal, instead of associational Fl´vio Code¸o a c Coelho information is mor intuitive and more reliable. Basic graph Causal relationships are a direct representations of our theory beliefs Bayesian Networks Direct representation of mechanisms Inference Simple to represent interventions thanks to modularity in the network Definition: Causal bayesian network Let P(v ) be a probability distribution on a set of V variables, and let Px (v ) denote the distributionresulting from the intervention do(X = x) that sets a subset X of variables to constants x.
    13. Causal Bayesian Networks Causal Bayesian Networks Fl´vio Code¸o a c Definition (cont.): Causal bayesian network Coelho Denote by P∗ the set of all interventional distributions Basic graph theory Px (v ), X ⊆ V , including P(v ), which represents no Bayesian intervention (i.e., X = ∅). A DAG G is said to be a causal Networks bayesian network compatible with P∗ if and only if the Inference following three conditions hold for every Px ∈ P∗ : 1 Px (v ) is Markov relative to G; 2 Px (vi ) = 1 for all Vi ∈ X whenever vi is consistent with X = x; 3 Px (vi | pai ) = P(vi | pai ) for all Vi ∈ X whenever pai is consistent with X = x.
    14. Causal Bayesian Networks Fl´vio Code¸o a c Coelho Basic graph theory Bayesian Networks Thank you! Inference

    + Flávio CoelhoFlávio Coelho, 2 years ago

    custom

    1111 views, 0 favs, 0 embeds more stats

    Introduction to Causal Bayesian Networks, based on more

    More info about this document

    © All Rights Reserved

    Go to text version

    • Total Views 1111
      • 1111 on SlideShare
      • 0 from embeds
    • Comments 0
    • Favorites 0
    • Downloads 14
    Most viewed embeds

    more

    All embeds

    less

    Flagged as inappropriate Flag as inappropriate
    Flag as inappropriate

    Select your reason for flagging this presentation as inappropriate. If needed, use the feedback form to let us know more details.

    Cancel
    File a copyright complaint
    Having problems? Go to our helpdesk?

    Categories