Causality and (Graph) Neural Networks

Поділитися
Вставка
  • Опубліковано 18 гру 2024

КОМЕНТАРІ • 20

  • @taranbarber5075
    @taranbarber5075 10 місяців тому

    The background information on causality was succinct and well put. Thank you!

  • @munozariasjm
    @munozariasjm 2 роки тому +1

    The contents you are creating are great!, keep them on

  • @qiguosun129
    @qiguosun129 2 роки тому +1

    Great lecture! Investigation of the combination of GNN and SVM may be promising in future research. Moreover, employing the technique to reinforce learning should also be interesting as it can introduce explainability to the black-box model.

  • @MelMashiku
    @MelMashiku Рік тому

    Thanks

  • @jordoobodi
    @jordoobodi Рік тому

    Causality can be biased in at least three ways: omitted variable, selection bias, reverse causality. Which one does GNN address?

  • @jenw62
    @jenw62 2 роки тому +1

    Love the content - especially the clarity and simple summary! Can you do a tutorial of a project using causal GNN framework? Can the GNN model first identify causal relationship? Then secondly, using that to create link between the node. Finally using GNN to predict an attribute of the nodes?

    • @jenw62
      @jenw62 2 роки тому

      If you can include comparison with other simpler causal inference model that will be great! (Model performance, explainability and training complexity)

    • @DeepFindr
      @DeepFindr  2 роки тому

      Thanks for the feedback and the suggestion! I add it to the list, but can't promise when I get to do it. Usually I set-up polls and let the majority decide :)

    • @VivekkannaJayaprakash
      @VivekkannaJayaprakash 2 роки тому

      @@DeepFindr Really appreciate if you decide to do the same.

  • @zhuangzhuanghe530
    @zhuangzhuanghe530 2 роки тому +1

    Is there any plan to publish a code tutorial?

    • @DeepFindr
      @DeepFindr  2 роки тому +1

      Hi! Not in the near future but I will add it to the list, thanks!

  • @yangbai657
    @yangbai657 2 роки тому

    I wonder if there is anything like more mainstream linear causality models like IV regression, difference in difference or regression discontinuity design with machine learning methods to incorporate nonlinearity in high dimensions.

    • @DeepFindr
      @DeepFindr  2 роки тому

      Good question :) I've not come across anything like that. But I'm pretty sure there are interesting ideas that can be applied in the machine learning domain.

    • @DeepFindr
      @DeepFindr  2 роки тому

      Maybe this paper: academic.oup.com/ectj/article/23/2/177/5722119

  • @zuu2051
    @zuu2051 2 роки тому

    Your Content is Super Amazing. I have a Kind Request. Can you please do a tutorial on Community Detection? That would be really helpful.

    • @DeepFindr
      @DeepFindr  2 роки тому

      Hi!
      Thanks :)
      Have you seen this Github summary? github.com/FanzhenLiu/Awesome-Deep-Community-Detection
      Besides that the next video is about unsupervised GNNs, which are also part of the methods used for community detection. Hope this will be benefitial :)

  • @hassaannaeem4374
    @hassaannaeem4374 2 роки тому

    great vid

  • @buddhidhananjaya8556
    @buddhidhananjaya8556 2 роки тому

    Hey, A quick question; would it be possible for you upload a video on crime forecasting using GNNs? There are practically no videos on UA-cam, we'll able to learn a lot especially about the MAPPING THE LOCATION DATA in a real world scenario! Datasets are already available online!