22 Matching Annotations
  1. Jul 2018
  2. Jun 2018
  3. Apr 2018
    1. The architecture and the network weights are trained and opti-mized using a score-based approach

      scores are derived via MMD and the gaussian kernel which is differentiable. So the loss function on our neural networks actually changes.

    2. Eis set to theuniform distribution on[0;1],U[0;1]

      Seems like they use the uniform distribution for error terms.

    3. nterventions

      interventions = freeze one variable and get join distribution from the rest.

    4. MaximumMean Discrepancy (MMD)

      MMD is given as the distance metric between \(P\) and \(\hat{P}\). MMD requires a feature mapping and uses gaussian kernel here.

      Here is a post on MMD: stats.stackexchange MMD

    5. Theideal score, to be minimized, is the distance between the jointdistributionPassociated with the ground truth FCM, andthe joint distributionbPdefined by the estimated(^G;^f;E)

      Score for these models given by the distance between the true joint distribution P and \(\hat{P}\). But what is distance?

    6. compact suppor

      A function has compact support if it is zero outside of a compact set. A subset of Euclidean space is called compact if it is closed and bounded.

      Basically we're saying there are functions f_i for deriving variable X_i from their parents.

    7. causal Markov

      all d-separations imply conditional independences in the observational distribution.

    8. causal sufficiency

      causal sufficiency = no unobserved confounders.

    9. skeleton

      skeleton is the undirected version of a directed graph.

    10. Causal generative neural networks - Use neural networks to as factors in functional causal model.

  4. Mar 2018
    1. The ontologies used to annotate histopathology findings (ie, theanatomy ontology and histopathology ontology) were storedcentrally within the eTOX project using a tool calledOntoBrowser

      annotations require ontologies. ontobrowser may be a good place to look.

    2. A massive curation effort took place to mapall verbatim terms used in reports to preferred terms from theontologies

      demonstrates a need for a curation platform.

  5. Jun 2015
    1. linkage analysis

      Analysis of genetic ancestry to determine likely regions of genetic disease etiology


  6. May 2015
    1. paper motivates the idea of michael's acceptor specific similarity map.